Load balancing mechanism for clustered PMIPv6 protocol

Load balancing mechanism for clustered PMIPv6 protocol Proxy Mobile IPv6 (PMIPv6) has become a credible member of pertinent research areas. This is attributed mainly to its capability of enabling mobility without imposing constraints or requirements on the mobile node (MN). This MN shield is enabled due to the transferring of mobility-related signaling to a new entity, which is called Mobile Access Gateway (MAG). However, associating MNs to a specific MAG inside the PMIPv6 network increases the MAG load probability. Thus, several research have enhanced the PMIPv6 protocol to improve its basic specifications and performance. Strategies include protocols, which apply the clustering technique to enhance the overall performance of the PMIPv6 in terms of routing, scalability, lifetime, and load balancing. The load balancing mechanism is considered in the non-clustered protocols. However, this mechanism has not been adopted in clustering-based protocols. Thus, pertaining to the load and the respective assignments is critical. In this article, to address these issues, a new load balancing mechanism is proposed among MAGs for Cluster-based Proxy Mobile IPv6 (CSPMIPv6) protocol. The signaling within the CSPMIPv6 has been enhanced to support the proposed load balancing mechanism. The proposed mechanism employs the inter- and intra-domain on a frequent basis to select the best MAG among the candidate MAGs. The new mechanism has improved the performance to create an evident improvement in terms of average queuing delay, handover latencies, transmission rate, end-to-end delay, and packet loss as compared to the LBM-PMIPv6 mechanism and CSPMIPv6 protocol. Keywords: Load balancing, Proxy mobile IPv6, IPv6 protocol, Queuing delay 1 Introduction the authentication, authorization, and accounting (AAA) In Mobile Internet Protocol (MIP), the high involve- server to register the MN with the respective LMA. The ment of MNs in the mobility-related signaling causes main role of LMA in the PMIPv6 protocol is to maintain several serious issues. Among the issues are long handover the MN accessibility whenever the MN changes its points latency and excessive signaling [1]. The MN is required of attachment within the PMIPv6 network. This removal to register with the home agent (HA) whenever the MN of responsibility from the MN results in the PMIPv6 changes its point of attachment. Addressing these prob- protocol enhancing the performance of MIPv6 protocol, lems associated with the MIP protocol, the Proxy Mobile especially in terms of traffic signaling, service disruption, IPv6 (PMIPv6) has been developed by the Internet Engi- and tunneling overhead. Therefore, making PMIPv6 a sig- neering Task Force (IETF) in order to effectively handoff nificant mobility management protocol for wireless sensor operations to MNs [2]. This is done by adding a new networks (WSNs). However, ignoring the load balancing entity, named Mobile Access Gateway (MAG) that takes among the MAGS and using single LMA to process or over the responsibility of mobility configuration from the forward the MN’s packets withing the LMA domain, have MN. The main role of the MAG entity is to detect MN resulted in many drawbacks (e.g, single point of failure, movement within the Local Mobility Anchor (LMA). In long handover latencies and intense signaling [3–5]). addition, the MAG initiates the required signals with To tackle these issues, research findings such as Sen- sor Proxy MIPv6 (SPMIPv6) [6–8], Cluster-based PMIPv6 *Correspondence: safwan_ghaleb@yahoo.com Department of Communication Technology and Network, Universiti Putra Malaysia, 43400 Serdang, Selengor D.E., Malaysia Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 2 of 23 for wireless mesh networks [9], andaCluster-basedProxy queuing delay. In the initial registration process of the Mobile IPv6 (CSPMIPv6) [4] have been developed to mit- MAGs and HMAGs, LB-CSPMIPv6 enables the LMA igate these problems. All these protocols have employed to assign a number for every sub-local domain in the clustering strategy in order to be more efficient for mobile clustered PMIPv6 domain. This domain is carried out users. The CSPMIPv6 [4]protocolsolvedahighnumber by the heartbeat message along with the load sta- tus in order to select the best MAG for the handoff of issues associated with the PMIPv6 and SPMIPv6 pro- tocols. Thus, the protocol is able to be used in a variety MN with the same domain of the MN’s serving MAG, of applications more compared to other protocols [3, 10]. which is different from existing schemes where each The CSPMIPv6 has inherited other drawbacks due to TMAG is selected based on its domain and load. In the dependency on the central and single LMA. The fast han- handoff process, LB-CPMIPv6 comprehensively consid- dovers for Proxy MIPv6 (PFMIPv6) [11]protocolhasbeen ers the scenarios of intra- and inter-handoff mobility to developed by the IETF to reduce the handover latency. provide a seamless mobility support to MHs roaming However, the serving network causes false handover initi- across various access networks, and low buffering cost, ation, due to the prediction of the target network to which which reduces handoff delay and prevents packet loss. theMNwillmove[12]. In this work, the CSPMIPv6 protocol handover signal- Contradictory to the benefits of the PMIPv6 proto- ing forms the core of the newly proposed load balancing col and its extensions, the constraints are caused by mechanism. The performance analysis of the proposed the MNs, which have to connect to a particular MAG load balancing mechanism with an extensive simulation within the PMIPv6 network. This causes the MAG to has been developed using Network Simulator (NS2) to be overloaded, especially in large networks. The over- show that the proposed load balancing mechanism (LB- loaded MAG causes a queuing delay, which in turn leads CSPMIPv6) achieves an improved quality of service (QoS) performance degradation packet loss, end-to-end delay, demands. and the throughput. There has been no consideration of In this work, the unique adoption of a load balancing load balancing in the basic specification of the PMIPv6 mechanism is developed to improve the overall system and its extensions. Thus, many types of research such performance of clustered PMIPv6 domain. as [13–18]haveattempted to solvethisissuethrough The main contributions of this article are as follows: applying the load sharing mechanism between the MAGs. 1. A detailed analysis of the CSPMIPv6 protocol in This has seen the increasing performance of the overall terms of merits, demerits, and its architecture, which system. Their proposed mechanisms, which are elabo- represents the underlying of the LB-CPMIPv6 rated in Section 3, deployed load balancing action by mechanism, is presented. The benchmark that has selecting the best target MAG, in addition to select- been selected for comparison purpose is reviewed ing the low-priority traffic MNs for the handoff process. extensively. These protocols have achieved good results in terms of 2. Providing an extensive overview of proposed striking a balance of load between the MAGs. How- mechanisms within the PMIPv6 domain. ever, these proposed mechanisms are applied only to 3. The development of a new load balancing non-clustered protocols. The clustering-based protocols mechanism for clustered PMIPv6 enhances the load have not researched, despite their widely being used in distribution among the MAGs within the CSPMIPv6 the research areas. Several issues such as high queu- domain. The focal point in this new mechanism is ing delay, end-to-end delay, and packet loss are accused exploiting the clustering benefits inside the PMIPv6 through applying these mechanisms when no consider- domain to enhance the process of selecting the ation is given to the division of clusters. Subsequently, TMAG during the handoff action. this is leading to serious disruption. As a result, it is evi- dent that the MAG selection has enormous potential for This article is organized as follows: enhancement, which is the focus of this article. The abil- Section 2 presents an extensive review of the CSPMIPv6 ity for serving network to select the Target MAG (TMAG) protocol focusing on its advantages, disadvantage, and in according to its domain will definitely lead to the reduc- particular the handover signaling. Section 3 deliberates in tion of the handover latency, end-to-end delay, and the detail the related work on the loading balancing in the average queuing delay. This is the result of the reduction PMIPv6 protocol. Section 4 discusses in detail the pro- of the signaling registration and the avoidance of the LMA posed LB-CPMIPv6 mechanism. In Section 5, a detailed involvement. explanation of the load balancing signaling for the clus- In order to achieve these potentials to increase the per- tered PMIPv6 domain is done and followed by Section 6, formance of the system, a load balancing based on the where the system architecture that is used as the environ- clustered PMIPv6 protocol is proposed LB-CSPMIPv6 to ment for the LB-CPMIPv6 mechanism is presented and provide a seamless mobility management and lowering the performance evaluation for LB-CPMIPv6 mechanism Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 3 of 23 is discussed. Section 7 concludes the contributions of the from the LMA, in order to mitigate the load and the proposed work. signaling on the LMA. In addition, the (AAA) func- tionalities are provided by the HMAG to reduce the 2 An overview of the CSPMIPv6 protocol registration time that is needed to register the MN. In this section, an extensive description of the CSPMIPV6 The registration processes of the new MN in the CSP- protocol, which has been used as a basis for the LB- MIPv6 protocol are performed according to the following steps: CPMIPv6 mechanism, is presented. Jabir et al. [4] proposes the clustered PMIPv6 archi- tecturetoovercomeproblemsassociatedwiththeProxy 1. Once the movement of MN has been detected by MIPv6 (SPMIPv6) [6] and Proxy Mobile IPv6 (PMIPv6) MAG , it sends a request message authentication to [2] protocols respectively. In this developed solution, the the AAA server including the MN identifier (MN-ID). PMIPv6 domain was divided into local sub-domains, 2. Then the MAG registers the MN in its domain as shown in Fig. 1. Each sub-domain contains several cluster by sending a Local Proxy Binding Update MAG clusters and each cluster is controlled and man- (LPBU) to the HMAG . aged by a cluster Head MAG (HMAG). As deliber- Upon the successful authentication by the HMAG , ated in the earlier sections, the CSPMIPv6 is derived the HMAG registers the MN on the LMA by sending from the PMIPv6, so functionalities of entities such as a Proxy Binding Update (PBU) including the MN-ID LMA, MAG, MN, and corresponding node (CN) are and the HMAG-ID. identical to those in PMIPv6 protocol. The new entity 3. Once the PBU message is received successfully by the HMAG in the CSPMIPv6 protocol has been configured LMA, a new Binding Cash Entry (BCE) is created to to take the responsibility of the local cluster handoff store the MN-ID and HMAG identifier. Internet CN LMA HMAG1 HMAG2 MAG1 MAG2 MAG3 MAG4 Intra-cluster Handoff Inter-cluster Handoff Fig. 1 Overall CSPMIPv6 system architecture Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 4 of 23 Subsequently, the LMA sends a Proxy Binding domain of the serving MAG. Thus, when the destination Acknowledgment (PBA) reply to the HMAG .The MAGacceptstoregisterthe MN,anLPBUmessage issent PBA message includes the Home Network Prefix to the respective HMAG including the MN-ID. However, (HNP) of the MN, which hereafter will be used for the HMAG will not find the MN-ID in its binding table maintaining the MN reachability within the PMIPv6 as the MN comes from another cluster. Therefore, the LMA must be involved in the process. This is done by the domain. The LMA configures the routing path with the HMAG by setting a bi-directional tunnel between requesting HMAG sending a PBU message to the LMA them to send and receive the traffic. advertising the new location of the MN. Subsequently, the LMA will update its BCE tables and send a reply to the 4. The HMAG adds the MN information to its Binding requesting HMAG. Update List (BUL) in order to register the MN and Once the requesting HMAG receives the PBA, a new sends a Local Proxy Binding Acknowledgment binding table for the MN will be created and a reply will (LPBA) message to the MAG containing the MN also be sent to the respective MAG. Finally, a new entry prefix. Then, routing configuration is performed to for the MN will be created by the MAG in its binding table make the MN accessible. and an HNP message will be sent to the MN. 5. When MAG gets the LPBA message from the The CSPMIPv6 has gained several substantial benefits HMAG , its BUL will be modified by adding the MN as a result of dividing the PMIPv6 domain into sub-local and forward the HNP to the MN through the networks. These advantages have increased the MN user advertisement message. Now, the MN has the ability performance concerning the mobility management. This to send and receive traffic. performance enhancement comes as a result of reduc- The MN information at the end of registration will be ing the LMA load by relieving it from the local mobility stored in the MAG ,HMAG ,andLMAtablesasstatedin i j signaling within the HMAG cluster. Furthermore, the sig- the aforementioned registration operation. Furthermore, naling cost has been reduced as a result of integrating The HMAG exchanges the MN information with the the AAA functionalities with the HMAG. Another critical MAG to perform a routing configuration for the MN. benefit is shortening the routing path when the MN moves Thus, there will be no need for a bi-directional tunnel inside its cluster (i.e., intra-cluster handoff) while per- set up between the HMAG and MAG [19]. Moreover, j i forming the handoff process by the HMAG without any the idea of integrating the AAA functionalities with the involvement from the LMA. Despite all of these merits LMA functions proposed by [6]isreusedinthisCSP- mentioned above, the CSPMIPv6 still suffers from sev- MIPv6 protocol to reduce the signaling cost during the eral issues such as the one point of failure (single LMA), MN registration. end-to-end delay, and excessive signaling [3]. The handoff procedure within the CSPMIPv6 domain functions is illustrated in Fig. 2 and deliberated as follows. 3 Related works on PMIPv6 protocol load When the MN decides to move from its serving network balancing mechanisms to another within the CSPMIPv6 domain, the MN move- In the PMIPv6, the mobility-related signaling responsibil- ments could be either an intra- or inter-cluster handoff. ity is undertaken by the MAGs on behalf of the MN. All In the intra-cluster handoff, the MN is supposed to move the MNs must be connected to a particular MAG which to another MAG within the same cluster domain. In other makes the MAG overloaded easily. The overloading on the words, the MN movement is still controlled by the same MAG leads to an increase of packet loss, end-to-end delay, cluster head HMAG. Therefore, the handover here is per- and the decrease of the transmission rate. Consequently, formed by the HMAG through updating its binding table several works have been proposed to reduce the load on without any intervention from the LMA. To do so, the the overloaded MAGs via applying the load sharing mech- destination MAG to which the MN decides to move, will anism between the MAGs to avoid the negative effect on send an LPBU message to the respective HMAG includ- the overall system performance. ing the MN-ID. Here, the HMAG will only need to update Kim and Lee [14] propose a load-balancing mechanism its table by setting the new MAG address in its MAG field to equitably distribute the load among different MAGs as opposed to the inter-cluster handoff. This is done once within the PMIPv6 domain. The proposed work led to the MN information has already been recorded. Then, the improving the overall system performance in terms of respective HMAG sends back an LPBA message, includ- average queuing delay, packet loss, and end-to-end delay, ing the HNP to the requesting MAG as well as configures while increasing the transmission rate. The authors uti- the routing performed with the requesting MAG in order lized the heartbeat message in order to allow for a specific to forward the MN packets. MAG to learn the load status of its neighboring MAGs. In the inter-cluster handoff, the MN movement is The heartbeat message is modified in order to store the detected by another MAG located outside the cluster load field for the load balancing action. Similarly, the Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 5 of 23 S-MAG1 MAG3 LMA HMN MAG2 HMAG1 HM AG2 MN occurrence detected LPBU message BUL Updating LPBA message BUL Updating MN occurrence detected LPBU message PBU message PBA message Updates BUL LPBA message BUL Updates Router Advertisement (HNP) Fig. 2 Handoff procedure in the CSPMIPv6 domain MAGs also sends a heartbeat message to the LMA, includ- reused in this work to create an identical platform for ing their load status. The LMA stores the received loads comparative purpose. in its BCE used to measure the overall system perfor- Another work in this area has been done by Kim and mance. The description of this is shown in Fig. 3.When Lee [13, 15] to enhance the load balancing by utilizing the LMA load exceeds a specific threshold, a heartbeat the IEEE 802.21 standard. The IEEE 802.21 optimizes message is sent by the LMA to the overloaded MAG. the handover between the heterogeneous technologies via Then, the overloaded MAG performs a load balancing facilitating media-independent handover by providing up and chooses the MNs that have the option to change layers with network-related information. This work aims their point of attachment. The target MAG is selected by to determine the load on a candidate point of attachment the serving MAG based on the received signal strength (PoA). There are cases where the PoA suffers from heavy (RSS) and the load status reported from the MNs. The loads as compared to the TMAG that experiences a lower signalingprocessthatis performedduringtheload bal- amount of load. This happens if the MAG load concen- ancing action is presented in Fig. 4.Thisworkrestricts trates on only one of its PoA (BS/AP). Thus, the target the procedure of choosing the handover MN (HMN) for PoA load is very important to knowing to reduce the over- the handover process by preventing the serving MAG to all load overhead. This proposed technique has proven to select the MNs that have a real-time session. Numerical have a remarkable enhancement in terms of queuing delay and simulation analysis has been conducted by the authors and transmission rate. to evaluate their proposed mechanism, and their result Another load balancing approach has been proposed shows significantly enhanced performance over the orig- by Kong et al. [16] for efficient migration of the load inal PMIPv6. The abovementioned mechanism forms the between the MAGs. Their approach determines the tar- core of the proposed LB-CPMIPv6 mechanism. Further- get MAG which needs a low signaling requirement. Each more, all the paper variables and assumptions are also MAG learns the load status of its neighboring MAGs by Inter-handoff-cluster Intra-handoff-cluster BCE Updating Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 6 of 23 MAG LMA Start Start Set initial values for Set initial values for ϴs ϴf, μ s,Nm ϴ ϴf No No On Packets On Packets receiving Receiving Yes Yes No No p >ϴ i i pt >ϴ Yes Yes No f<ϴf MAG preforms a load balancing action Yes MAG select the handover MNs (HMN) LMA sends load balancing request message to MAG MAGi selects the best MAG from the non- i overloaded candidate MAGs MAG obtains the MAGs load from the Heartbeat message exchanges among the MAGs MAG sends handover command to HMNs to change their routing to the TMAG MAG sends load status to LMA periodically Fig. 3 Load balancing operation in the PMIPv6 domain [13] exchanging their load among each other in the domain. makes the MN moves to a different domain that requires Then, the MAGs create a list of candidate MAGs based extra signaling, which in turn leads to high queuing delay on the received load information in order to select the and low transmission rate. best TMAG for the HMN. A proactive load balancing is An agent-based scheme was proposed by Dimple and performed during the initial attachment of the MN by Kailash [17] to mitigate the overloaded MAG issue within selecting the MAG that has the lowest load according the PMIPv6 network. Their mechanism works by mov- to the load information. This is done before the current ing the mobile agent from one location to another to MAG becomes overloaded. Therefore, by avoiding the reduce the load on the overloaded MAG. The mobile overloaded MAG, benefits such as low packet loss and agent achieves this through visiting one MN to collect its low signaling will be achieved. However, in this mecha- data and moves to the other MNs associated to the MAG nism, the HMN experiences an extra delay, especially in to take the only relevant data for transmission, in order to the proactive scenario caused by the time needed by the reduce the overhead communication. The MN selection is serving MAG to determine the best TMAG to which the performed according to certain criteria such that the MNs that have real-time session will not be selected, while the HMN moves according to the MAGs loads. Real-time ses- sions have not been considered in [16], which in turn MNs that have high-rate data connection become a target degrades the system performance. Moreover, this mecha- for a handoff. Despite the benefits gained by employing nism requires MNs with multi-interface to be connected the MN agent, several issues arise. Anticipating the MN with two different networks, which makes it restricted in the load balancing adds some burden to the MNs and to this scenario. Also, multi-domains within the same increase the function complicity. This is done by selecting domain has not been considered in this work, which one MN to visit the other MNs within the MAG domain Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 7 of 23 HM N LMA SMAG C-MAG1 C-MAG2 Report (MN-ID, MN-IID, RSS, New MAG Info.) HI Message HI Message Determine the MAGs Load HAck Message (C-MAG1 load Info. HAck Message (C-MAG2 load Info. Determine the TMAGs for the HMN Handover Command(TMAG-Info.) Establish L2 connection by the MN MN Starts New Connection PBU Message PBA Message Router Solicitation (RS) Router Advertisement (RA) Fig. 4 Load balancing signaling in the PMIPv6 domain to collect the similar data packets which require some sig- that satisfy the mobile users. However, employing the naling messages between the MN and the associated MN. Global Position System (GPS) expedites the power of Moreover, the employed θ threshold by the LMA in the the MNs, which is not acceptable in the critical applica- mechanism depends on the size of the data reduction by tions . Besides, this work consecrates on the overloaded the MAG that sent to the LMA. This leads to overload MAGs and ignore the overloaded LMA. The overloaded the MAG that has numerous attached MNs but does not LMA is determined according to the all MAGs load have any similar data between them or have less than the in the system, which may be accrued even when the specified threshold, which not reflect the reality load state MAGs are not overloaded. This definitely degrades the of the MAG. Furthermore, clustered PMIPv6 protocol not overall system performance through increasing the time considered in their implementation, which in turn may of registering/de-registering the MNs (large queuing lead to effect the intra-domain mobility advantages in a delay). Furthermore, divided domains do not consider in contrary manner. their work, which affects the QoS regarding the mobility Qutub and Anjali [18] introduce an efficient mechanism management. The load balancing problem also has been researched by to balance the load among the overloaded and low-load MAGs. Their mechanism selects the target MAG accord- the Internet Engineering Task Force (IETF) and for which ing to its geographical serving area and its current load. a Request for Comments (RFC) was introduced by Jiang Also, the MN selection for handover is performed based [20]. Each MAG sends its load periodically to the LMA on the MN’s QoS profile, location, direction, and multi- and hereafter is used by the LMA to create a list of can- interface capability. This selection has proven to reduce didate MAGs for performing load balancing. The factors the overload and provide the service, which satisfies the have been used in their mechanism to select the target QoS. In this work, not only the overloaded is avoided MAG are specified in [18]. The process of selecting the but also the services are provided with a level of QoS HMN is performed as follows: Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 8 of 23 1. When the MAG becomes overloaded, the MAG these sessions according to same criteria as specified in starts load balancing action by selecting the HMN [21]. Their experiments and numerical results that have according to its service type to avoid selecting the been conducted have shown significant improvements in MN that has a real-time session. terms of load distribution as well as reducing the multicast service disruption. However, moving the MN’s multicast 2. Then, the MAG sends Load State Message (LSM) to session to another LMA (LMA has least load), which may the LMA in order to inform the LMA about its loads. 3. Accordingly, the LMA gives a feedback to the MAG be located far away from the MN affects the system per- about the overloaded MAGs and the non-overloaded formance. This is due to the long path between the MN MAGs. This is done by sending a Load State and the new LMA that leads to long delay or packets Acknowledgment Message (LSAM) to the MAG in drops high signaling cost. Moreover, this work focuses on order to migrate the HMN to a new MAG that is not balancing the load between the LMAs and ignores the overloaded. overloaded MAGs, which may overload even when the load on LMAs is balanced. 4. Once the MAG receives the LSAM message, it sends Another work focuses on distributing the load between a request message to non-overloaded MAGs. the LMAs introduced in [23]. The primary aim of this 5. The non-overloaded MAGs upon receiving the work is moving the load from the overloaded LMA to request messages reply to the requested MAG along the LMA has the least load. This is done when the load with the acceptance or the rejection of its request at LMA reached the specified threshold. Accordingly, the according to their status. LMA sends load balancing (LB) warning to the MAG that 6. Then, the overloaded MAG sends a notification to serves the selected MN. Then, the MAG sends refresh the HMN including the information about the binding to the LMA and hereafter the LMA communi- TMAG. cates with the new LMA to bind the selected MN to 7. The MN once receives the notification from the another LMA. Now, the MN anchored at the new LMA. MAG, it sends Router Solicitation (RS) message to This work shows remarkable improvements regarding the TMAG to inform it about its movements. the blocking probability and dropping probability than 8. A PBU and PBA messages are exchanged between PMIPv6 with no load control. However, the authors do the MAG and the LMA to register the handoff MN. not take into consideration the overloaded MAG, which 9. Finally, the TMAG sends a Router Advertisement is easily susceptible to be overloaded any time when the (RA) message to the HMN including the new IP attached MNs very high or when the MN requires a high addresses in order to complete its registration. stream session. This leads to service disruption through According to the registration procedures in this mech- increasing the queuing delay. In addition, the hierarchical anism, the MAG should send a request to the all non- domain is not considered also in their work, which may overloaded MAGs and await their responses to select the lead to high queuing delay. TMAG for the HMN based these responses. This process A load balancing scheme is introduced in [24]to consumes the bandwidth due to the messages exchanged improve the overall system performance in terms of between the MAGS during the load balancing mecha- handoff delay and throughput. The IEEE 802.21 Media nism. In addition, this work does not target the overloaded Independent Handover Services (MIH) functionalities LMA, which is responsible for the acceptability of all the are utilized with the proper selection of the MN new MNs connected to the MAGs that may be overloaded. network to provide a seamless handover in the heteroge- Moreover, the intra domain mobility in case of divid- neous networks. In this scheme, when the signal of the ing the PMIPv6 into sub-local domains is not considered, MN becomes very weak, a report from MN is sent to the which lead to long handoff delay through increasing the MN serving MAG. Then, the PMAG upon receiving the path recovery and signaling cost. report sends handover initiate (HI) message to the LMA Nguyen and Bonnet [21, 22] introduce a solution mech- including all candidates MAG/APs information. Accord- anism to solve the issue of load balancing in the PMIPv6 ingly, the LMA forwards the HI message to the candidates by considering the IP multicast session. Their solution and these candidates response to the LMA with sending a caters two scenarios, which are named as proactive- Handover Acknowledgment (HAck) messages to inform multicast and the reactive-multicast. For the former, when the LMA about their status and their acceptance to serve the MN. The LMA forwards the received HAck messages the MN starts a new multicast connection, a load balanc- ing action will be triggered to select the suitable LMA to the serving MAG in order to select the proper network to manage this connection. However, in the latter, when for the MN. Despite the enhancements that are made in the LMA becomes overloaded, the LMA starts to select this scheme regarding the handover time and throughput, some of the multicast sessions for a load balancing pur- additional signaling messages are required between the poses. Then, the LMA selects the suitable target LMA for PMAG, LMA, and the candidates MAGs/APs, which Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 9 of 23 negatively impact the system. The reactive load balancing issues focused on the clustered protocol and to provide is not considered in this scheme, which leads to increas- solutions. ing the blocking probability in the overloaded MAGs and In this work, the CSPMIPv6 protocol is implemented to service disruption due to increasing the queuing delay in make it as the central referral platform for the proposed the overloaded MAGs, in addition to ignoring the divided LB-CPMIPv6 mechanism. The proposed mechanism in [13] has been implemented in this research work whereas domain as the previous works do. Raza et al. [25] employ the Software Defined Network- it is applied on the CSPMIPv6 for a comparison pur- ing (SDN)-based solution in order to mitigate the loads pose. The CSPMIPv6 architecture is shown in Fig. 1.This between the LMAs. This works depend on central mobil- protocol divides the PMIPv6 domain into sub-domains. ity controller that is responsible for monitoring the load at Each domain encompasses some MAGs that form a clus- the LMAs. The controller upon detecting the load cross- ter within the PMIPv6 domain. Subsequently, each cluster ing over the predefined threshold on any of the LMA elects one MAG to act as the cluster head (HMAG). starts moving some traffic from the massive LMA load to The MAG in the CSPMIPv6 can be easily overloaded as the lower LMA load. According to the analytical analysis, in PMIPv6 protocol. Figure 5 shows an example of the their scheme has significant improvements regarding dis- CSPMIPv6-based inter-architecture of its overlapped area ruption period of uplink and downlink traffic during load among its clusters. The overlapped area between the sub- balancing action compared to their benchmark. How- domains contains a number of HMN candidates. These ever, adding extra element is costly. In addition, the mas- candidates must have another optional network to con- sive MAG load is not considered in their scheme, which nect with for the handover purpose. As seen in Fig. 5, affects the system performance in terms of handover MAG1 andMAG2are locatedinthe sameclusterasthe delay. Moreover, LMA domains also not consider in this HMAG1, while MAG3 is located in a different cluster scheme, which leads to moving the track to another LMA HMAG2. The solid lines represent the current connection located far away from the serving LMA. Furthermore, of the HMN candidate, while the dotted lines represent scalability issue has arisen as a result of using a central the optional connection for it. The selection process of controller. TMAG and HMNs must be performed to provide bet- SDN also used by [26] to reduce the blocking proba- ter performance to the HMNs. Likewise, choosing the appropriatenetwork fortheHMNsin theoverlapped bility and increase the resource utilization through using area leads to the balance of load between the MAGs, mobility-aware load distribution for multiple controllers. The objective of this work is handling the handover which in turn avoids the overlapped MAGs. The proposed messages as fast as possible. This is performed by dis- enhanced load balancing algorithm is presented in the tinguishing the handover messages (gives them high pri- next section. ority) and manipulate them by the controller has the least load among the other controllers if the serving 4 The proposed LB-CPMIPv6 mechanism controller suffers from heavy load. However, the main In this paper, a new mechanism, named LB-CPMIPv6 consideration is given to the loads on the LMA and is is proposed to enhance the overall system performance ignored the loads on the MAGs. In addition, the clus- of IP-WSNs by considering a load balancing approach tered domain also is not considered in their scheme. These in the clustered network. The proposed LB-CPMIPv6 ignoring lead to serious issues regarding the mobility, mechanism expands the MAG capability to avoid over- which in turn affect the service delivered to the mobile loading issue by developing a new load balancing mech- users. anism. In addition, the proposed mechanism reduces the The review of these deliberated algorithms raises some time needed to recover path between the communicating majorconcernswhichhavetobeconsideredfor theload entities. sharing mechanism. A list of candidate MAGs to be cre- In the proposed LB-CPMIPv6 mechanism, a domain ated with a fewer message exchange to avoid the network number should be assigned to every sub-domain in overloading issue and choosing the HMNs should be order to distinguish between the clusters within the performed based on their traffic type to avoid the selec- PMIPv6 domain. The proposed LB-CPMIPv6 mecha- tion of the HMNs that have an arguing critical-session. nism provides an efficient way to balance the load Unfortunately, proposed works above metioned have not between the MAGs, by predicting the proper TMAG proposed such solution for the clustered-based protocol to which HMN moves accurately, as illustrated in Algo- during the formation of the candidate MAG list. Thus, is rithms 1, 2, and 3. Algorithms 1, 2, and 3 explain effected the overall system performance as the selection the functionalities of MAG, HMAG, and LMA respec- of TMAG from another cluster or in the case, there is tively within the proposed mechanism. The control another target MAG from the same cluster of the serving flow diagram of LB-CSPMIPv6 mechanism is illustrated MAG. Thus, this work has been motivated by these open in Fig. 6. Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 10 of 23 MAG MAG MAG1 MAG3 HMN HMAG 2 MAG HMAG 1 MAG2 MAG MAG MAG MN device HMAG Fig. 5 An example of CSPMIPv6 inter-structure for load balancing movement 4.1 Load balancing mechanism for clustered PMIPv6 domain (LB-CPMIPv6) In this proposed mechanism, a load balancing mechanism is developed for a cluster PMIPv6 to improve the efficiency of MNs and accordingly the overall system Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 11 of 23 MAGi HMAGk LMA Start Start Start Set initial values for Set initial values for Set initial values for s f, µ f , s,Nm , L, fL No No No On Packets On Packets On Packets Receiving receiving Receiving Yes Yes Yes No No No pi > i pt > pT > Yes Yes Yes No No f< f fL< fL MAGi preforms a load balancing action Yes Yes MAG select the handover MNs (HMN) HMAG sends load balancing request message LMA sends load balancing request message to MAGi to HMAG MAG selects the best MAG from the non- overloaded candidate MAGs MAG obtains the MAGs load from the Heartbeat message exchanges among the MAGs MAGi sends handover command to HMNs to change their routing to the TMAG MAG sends load status to HMAG periodically i K HMAG sends load status to LMA periodically Fig. 6 Load balancing operations within CSPMIPv6 domain performance is improved. This is due to the need to and will hereafter be denoted by λ where the N is the i m take into consideration the intra- and inter-domain mobil- number of the measurements. The N measurements are ity during the load balancing process. The MAG located used to estimate the λ for MAG , which is computed as within the CSPMIPv6 domain acts as the gateway between theaverage arrivalrateatacertaintimeinterval. After the MNs and the HMAG. Thus, the MNs must be con- that, the MAG calculates the average packet arrival rate nected to the MAG to be connected to the network. using the weighted moving average technique under the Subsequently, the MAG could become overloaded if the assumption that μ istheaverage servicerateoftheMAG , i i number of the connected MNs increases in the net- whichisusedby[13] and is mathematicaly expressed as work. This constraint has motivated, a new load balancing follows: mechanism, which is applied to reduce the load at mainly the overloaded MAGs. (N − j + 1)λ m N −j+1 j=1 λ = (1) In order to ensure the standardization of the per- j=1 formance analysis as a comparative platform, the LB- CPMIPv6 mechanism performance analysis has reused Thereasontoutilizetheweightedmovingaverage the parameter values and assumptions that have been pre- method is to reveal the uncontrolled action. In addition, sented byKimand Leein[13]. Hereafter, the proposed it gives a higher weight to the current traffic sample as load balancing mechanism for the PMIPv6 network in [13] compared to the old traffic sample in the measurement as should be referred as the “LBM-PMIPv6 mechanism” for proposed in [27]inorderto computetheMAGloadpre- the sake of simplicity. In the CSPMIPv6 system model, the load at MAG is cisely. Then, the p can be expressed as where the λ is i i measured according to the average packet arrival rate in a theaverage arrivalrateand the μ is the average services particular interval time. The similar measurement is used rate at a certain time. By considering the MAG process- th for measuring the arriving rate at a certain j time interval ing capacity into account, θ is the maximum acceptable i Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 12 of 23 load on the MAG i.If p >θ ,MAG becomes overloaded the HMAG will send a heartbeat message to its related i i i k and will perform a load balancing technique. MAGs to inform the most overloaded MAGs to perform In the LB-CPMIPv6 mechanism, every MAG is sup- a load balancing action. In this work, the extended field (F posed to send its load and its domain number in a flag and load status field ) in the heartbeat message that is periodical manner to the related HMAG using the heart- given by [13]isreusedinthe sameway. beat message, which has been introduced in [28]. The Finally, every HMAG in the CSPMIPv6 domain has to heartbeat message is exchanged periodically by the MAG send its load that is received from the related MAGs to the information related to their related HMAGs within the LMA. This is performed using the heartbeat message to CSPMIPv6 domain. This is done to inform the HMAGs compute the overall system performance and operates as with their loads as well as to detect the reachability of follows. the other end links. In this work, the heartbeat message Once the HMAG receives the total loads p from each is extended to include the load status and the domain related MAGs, the HMAG send these loads to the related number, as shown in Fig. 7. In addition, the Proxy Bind- LMA periodically using the heartbeat message. ing Acknowledgment (PBA) and the Local Proxy Binding Upon receiving the p loads from the associated Acknowledgment (LPBA) messages, which are presented HMAGs, the LMA computes the load status F1, using in [2]and[4] respectively are extended, as shown in the received loads to measure the overall system perfor- Figs. 8 and 9. This extension is to include the domain mance within the CSPMIPv6 domain. The total load can number during the initial attachment of the MAGs while be expressed as: in the other control signaling the domain number is set to be zero. These two mechanisms are used to define p = p T t the domain numbers, which are dynamically done by the k=1 LMA according to the number of its related domains or statically done during the installation. where the N isthenumberoftheHMAG in theCSP- MIPv6 domain and p is the total load at the HMAG . t k i=1 After that, the LMA in the CSPMIPv6 domain measures f = (2) 2 the F1 according to the MAGs load received by the related M ∗ ( p ) i=1 HMAG in the entire system as expressed in Eq. (2), where Each domain number and loading information that is M denotes the number of HMAGs in the system. received by the HMAGs, LMA, and MAG are saved If f is less than θf , a heartbeat message request is sent in their databases, which are then used in the intra- by the LMA to the most overloaded HMAGs to inform and inter-domain load balancing. Subsequently, every them to perform a load balancing action. The LMA uses HMAG also measures its load status, F1, via employing k the received loads from all the related HMAGs to deter- the stored load information on its database that is received mine the most overloaded HMAGs. A new flag is also from its related MAGs within its domain. The total load added to the heartbeat message is named F andisset to1if for each HMAG domain p can be expressed as p = t t the heartbeat message comes from the LMA entity to the p where the M is the number of the MAGs within related HMAGsorzeroifthe heartbeatmessage comes i=1 the HMAG domain in the CSPMIPv6 domain. If p >θ, k t fromtheHMAGtotherelated MAGs. the HMAG measures the Fairness Index (F ) according to k I Once the overloaded HMAG receives the heartbeat the Eq. 2 that is given by [29]. The F lies between 0 and 1. message, a load balancing action is performed by sending If all the MAGs within the HMAG domain have the same a heartbeat message to the related MAGs, which in turn load, the F is 1. selects the HMNs in the overlapped area. The HMNs must Subsequently, the HMAG uses the MAGs load infor- change their point of attachment to another MAG. The mation, which is stored in its policy database to com- criteria of the HMN selection are discussed in the next pute the f value and the compares it with θf.If f <θf, subsection. 01 2 3 0 1 2 3 45 6789 0123456 7 8 9 012345 67 8 9 0 1 Reserved L F U R Sequence Number Heartbeat Time Interval Load Information Domain Number Fig. 7 Heartbeat message, including the domain number Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 13 of 23 01 2 4 0 1 2 3 4 5678901 23 4 5 6 7 8 9 01 2 3 456789 01 Sequence Number AHLKMRPSNQReserved Life Time Domain Number Mobility Options Fig. 8 PBA message, including the domain number According to the above explanation, the LB-CPMIPv6 After the HMNs selection by the MAG from its candi- mechanism only affects the protocols with multiple date list, the MAG now is ready to choose the best target domains. In other protocols, there is no impact because MAG to where the HMNs bind. The selection of the tar- the LB-CPMIPv6 mechanism deals with the whole get MAG is performed as described in the subsection of domain as one domain. Thus, selecting the MAG will be the target MAG selection. performed based on the LBM-PMIPv6 mechanism [13]. 4.1.2 Target MAG selection 4.1.1 HMN selection For an enhanced load distribution, the selection of the tar- The wrong selection procedure of the HMNs candidate get MAG must be performed as accurately as possible. significantly degrades the performance of the system. Therefore, the Technique for Order Preference by Sim- Thus, the system must be adapted to appropriately choose ilarity to Ideal Solution (TOPSIS) algorithm is modified the HMNs in a better manner. In this work, the process to determine the target MAG in this research. Further- of HMNs selection by the overloaded MAG is performed more, in this study, additional operations are proposed to based on some criteria as follows. adopt the algorithm with the clustered PMIPv6. The intra- The MAG chooses the HMN that has an option to and inter-cluster handoffs within the CSPMIPv6 domain change its connection. This indicates that the HMN is have been considered due to the adaptation process. The located between different networks that are advertising system has achieved better selection technique to the tar- their services to such HMN in order to maintain a contin- get MAG among the candidates MAGs, which have been uous IPv6 session. This is achieved if the HMN is located reflected in the system performance in terms of handover in the MAGs overlapped area and it receives a Signal latency, end-to-end delay and queuing delay as presented Strength (SS) from all of them. in Section 6. The enhanced processes of selecting the best The MAG creates a candidate list for the HMNs that target MAG are performed as follows: exist in its overlapped and receive an RSS from another The MAG utilizes the RSS, which is reported by the MAG. HMN to determine the available network for the HMN. Then, the MAG should select the HMNs that require The MAG then places the available networks as candi- the highest data rate from the candidate HMNs. date networks (i.e., candidate MAGs). This is performed TheMAGshouldnotselectthe HMNsthat havea to select the best candidate MAG in terms of RSS, load real-time connection (e.g., audio and video) due to the status and the domain number during the load balanc- sensitivity to delay leading to increase the handoff latency. ing action. The technique for order preference is adapted To determine the non-real HMN session by the MAG, from the TOPSIS algorithm that is presented in [27]. The the “Traffic Class” or the “Flow Label” field of IPv6 must TOPSIS algorithm is used by the serving MAG to deter- be examined by the MAG for all the candidate HMNs mine the target MAG. This algorithm used to choose the in the MAG overlapped area [30]. So disturbing the real- optimal MAG as possible according to the Signal Strength time session during the handover latency will be avoided. reported by the HMNs and the load status of that MAG. If the MAG overlapped area does not contain any non-real Observation shows that the TOPSIS algorithm is not suit- HMN session, the HMNs with the highest data rate will be able when applied within the clustered PMIPv6 domain. selected. This is due to the fact of dividing the domains into local 01 2 3 0 1 2 345 6789 0 1 2345 67 8 9 0 1 2345 67890 1 Status K R P S N Q Reserved Sequence # Life Time Domain Number Mobility Options Fig. 9 LPBA message, including the domain number Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 14 of 23 sub-domains, which is not considered in the TOPSIS with the normalized signal strength. For the sake of sim- algorithm, subsequently increasing the communication plicity the wp ¯ is replaced by v ,wherethe wS is replaced i p i overhead. Therefore, some enhancement has been imple- by v mented by the LB-CPMIPv6 mechanism starting with Step 5: Determine the optimal network C (min v ,max modification of the exchanges of messages between the v and the worst network C (max v ,min v )from V S p S system entities until of change the HMN its points of matrix: attachment to ideal TMAG. The TOPSIS algorithm steps ∗ ∗ ∗ C = v , v C ={v ˆ , v ˆ }, ∀ ∈{1, 2, ... , n}.(9) including the additional operations of selecting the opti- p S i p S mal target MAG for the HMNs are described as follows: Step 6: The TOPSIS algorithm calculates the separation Step 1: The TOPSIS algorithm constructs the decision measures by using the Euclidean distance. The separation matrix D, which is a [1 × n] matrix, as: of each candidate MAG from the optimal and the worst D = {C , C , ... , C },(3) 1 2 n MAG, S and C are calculated as: th where C represents a pair of p and S for the i candidate i i i ∗ ∗ ∗ 2 S = v − v (v − v ) , ∀ ∈{1, 2, ... , n}. (10) p S i i p MAG in the matrix D,as: C = (p , S ) , ∀ ∈{1, 2, ... , n},(4) i i i i ˆ 2 S = v − v ˆ (v − v ˆ ) , ∀ ∈{1, 2, ... , n}. (11) i p p S S i where p denotes the load status and S represents the sig- i i nal strength of candidate MAG and n isthenumberof Step 7: Next, the MAG ranks the preference order after candidate MAGs in the matrix D. Moreover, the TOPSIS calculating the relative separation measure as: algorithm is designed to avoid the candidate MAGs, which ˆ ˆ load is more than a predefined threshold θ. x = S /S + S , ∀ ∈{1, 2, ... , n}. (12) i i i i Step 2: The TOPSIS algorithm computes the normalized Step 8: This step is performed according to two different decision matrix D, which is a [ 1 × n]matrix, as: cases. First, if the closest MAG to the ideal network envi- ¯ ¯ ¯ ¯ D = C , C , ... , C,(5) 1 2 n ronment has the same domain of the serving MAG, the serving MAG selects it without any hesitation as the tar- ¯ ¯ where C represents a pair of p ¯ and S for the each i i i get MAG. Second, if the closest MAG to the ideal network candidate MAG in the matrix D,as: is from a different domain, the serving network looks into ¯ ¯ C = p ¯ , S , ∀ ∈{1, 2, ... , n},(6) the ideal network candidate MAGs list to see if there is i i i i another ideal one has the same domain in order to choose it instead of the ideal one has a different domain. The where p ¯ = p ¯ / p ¯ and denotes the value of normalized i i i=1 selection of the TMAG from the ideal network candidates ¯ ¯ ¯ maintains the system stability in terms of signal strength load status, while the S = S / S and represents the i i i=1 and the load status. This means, the process of selecting value of normalized signal strength of candidate MAG the TMAG that have the same domain will not affect the and n is the number of candidate MAGs in the matrix D. thresholds of the signal strength and the load, which will Step 3: The TOPSIS algorithm uses the weight value (w), lead to the maintenance of the connection session without which is a system parameter. This weight (w) and its com- any service disruption. plement to 1 (1 − w) are used to weight the p and S , i i Step 9: When the serving MAG has done the selection of respectively. Since the load at the candidate MAG (p ) has the closest MAG to the ideal network environment from higher priority than the signal strength (S ), the weight the ranking preference, the serving MAG starts a load value (w) have to always be greater than 0.5. balancing process by sending a handover command mes- Step 4: The TOPSIS algorithm calculates the weighting sage to HMN. In order to select the ideal TMAG, a new decision matrix: V,as: load balancing signaling is proposed. In the next section, the new load balancing signaling within the clustered ¯ ¯ ¯ V = wC , wC , ... , wC,(7) 1 2 n protocols is explained in detail. ¯ ¯ where wC represents a pair of wp ¯ and wS for the each i i i candidate MAG in the matrix D,as: 5 Load balancing signaling for the clustered PMIPv6 domain ¯ ¯ wC = wp ¯ , wS , ∀ ∈{1, 2, ... , n},(8) i i i i As mentioned earlier, every MN within the PMIPv6 where the wp ¯ is the result of the multiplication of the domain must connect to a particular MAG to commu- weight value with the normalized load status value and the nicate with other MNs. This can lead to overloading the wS is the result of the multiplication of the weight value MAG, especially in large networks, when the number of i Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 15 of 23 MNs is substantial. This section presents the required sig- When the MAG receives the handover report from naling of our proposed work to mitigate the load of the the HMN, the MAG extracts the IDs of the candidate overloaded MAG. Figure 10 shows that this signaling is MAGs, which is given by the report and subsequently extended from the CSPMIPv6 signaling framework, which stored as a list on its policy database. Then, the MAG includes the inter- and intra-handover signaling opera- utilizes the Handover initiates (HI) and the Handover tion [4]. Given that, the MN sends a report to the serving Acknowledgment (HAck) messages, which are introduced network, which includes the MN-ID, MN-IID, the new by [11] and extended by [13] to determine the load of MAG-ID,andtheRSS.Thisreportissentonlyifthe RSS the candidate MAGs. This is done by sending an HI exceeds a threshold as performed in Third Generation message by the MAG to all candidates MAGs includ- Partnership Project (3GPP). The scenario of performing a ing the MN-ID. Once the HI message received by the load balancing handover in the clustered PMIPv6 domain candidate MAGs, the candidate MAGs reply by send- is performed as follows. ing HAck messages emulated to the MAG including The MAG performs a load balancing procedure accord- the current loads information. A flag N used as intro- ing to three cases. Therefore, when a load of MAG duced by [13] to distinguish between the HI and HAck exceeds a specific threshold or if its cluster head HMAG messages. sends a load balancing request or if the LMA orders The MAG employs the received loads information from the HMAG , to perform a load balancing, which in turn thecandidateMAGsandthedomainnumbertochoose moves this order to the related MAG .Afterthat,the the TMAG, which the HMN will move to as described MAG starts to perform a load balancing procedure by in Section 2.2. Subsequently, a Handover Command Mes- choosing theappropriateHMNfromtheoverlappedarea sage (HCM) is sent by the MAG to the HMN. to perform a handover action. This HMN selection is Upon receiving the HCM, the HMN starts a new link performed as presented earlier. connection with the TMAG. HMAG2 HMN SMAG HMAG1 C-MAG1 C-MAG2 LMA Report (MN-ID, MN-IID,RSS, New MAG Info.) Tunneling HI Message HI Message Determine the MAGs HAck Message (C-MAG1 Load Info. Load HAck Message (C-MAG2 Load Info. Determine the TMAGs for the HMN Handover Command(MAG1-Info.) Establish L2 connection by the MN MN Starts New L2 Connection LPBU Message LPBA Message Tunneling Router Solicitation (RS) Router Advertisement(RA) DATA Handover Command(MAG2)) Establish L2 Connection by The HMN MN Starts New L2 Connection LPBU Message PBU Message Updating BUL PBA Message LPBA Message Updating BUL Router Solicitation (RS) Tunneling Router Advertisement (RA) DATA Fig. 10 Load balancing signaling in the CSPMIPv6 domain Inter-cluster Loab balancing Intra-cluster Load balancing Extra Signaling for LB-CSPMIPv6 Updating BCE Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 16 of 23 Table 1 Parameters for experimental results When the TMAG detects the attachment of the HMN, Parameter Values an LPBU message sends to the HMAG . Once the HMAG receives the LPBU message, the k θ 0.925 HMAG fetches its binding table by searching for the θ 0.105 HMN information. If the HMN moves to another MAG θ 0.5 within the HMAG cluster domain, which is as called θ 0.5 intra-cluster handoff mobility, the HMAG updates its θ 0.175 fL Binding Cash Entry (BCE) by setting the New MAG (NMAG) address instead of the TMAG address in its μ 2 packet/ms MAG field. On the other hand, if the HMN moves to K 20 another MAG in a new HMAG (NHMAG) cluster domain Packet size 1024 KB within the CSPMIPv6 domain, which is as called inter- cluster handoff mobility, the NHMAG will not find any matching entry in its BUL table. Since the HMN comes from another cluster, so the NHMAG sends PBU to the the network simulator NS2 [32, 33]. As illustrated in LMAtoupdateits BCEfor theHMN.Then, theHMAG Fig. 11, which represents the system topology, the MNs field entry in the BCE of the LMA will be modified by communicate with each other including the CN that is setting the NHMAG address and releases the old one for connected to the CSPMIPv6 domain through the core net- the HMN. After that, the LMA replies to the NHMAG via work. The Poisson process has been used by CN and MNs sending a PBA message. to generate packets with a mean rate of 2 packets/ms. The NHMAG, upon receiving the PBA message sends Each MAG measures the packet arrival rate (λ)atevery an LPBA message to the NMAG after updating its binding 50 ms. When the number of measurements reaches 20 entry for the HMN. times by the MAG, the MAG calculates the λ using the The NMAG consequently updates its BUL for the HMN weighted average method as depicted in Eq. (1). In the and advertises the HNP to the HMN. proposed network topology, the MNs are randomly scat- tered over 10 MAGs. In addition, the MAGs are equally connected with two HMAGs (HMAG1, HMAG2) while 6 Performance evaluation the HMAGs are associated to one LMA. The load status This article presents the development of a load balancing foreachMAGiscarried by theheartbeat message thatis mechanism among MAGs in the CSPMIPv6 domain. The sent every single second to their related HMAG. The time performance analysis of this algorithm and the compara- between the first Heartbeat and the next should be small tive analysis has been done using discrete event simulation as recommended in [28]. Similarly, the HMAGs send their and in particular the NS2. For the evaluation purposes, load status to the LMA to measure the overall system per- the work in [13] is re-implemented and the CSPMIPv6 formance. Several MNs are scattered randomly over the protocol, which is presented in [4]. Also, this work reuses MAGs areas. The total load varies between 0.05 to 0.8. the parameter values and assumptions that have been Every MAG is associated with a limited queue K as men- used in [13] to ensure a level of comparative platform, tioned in Table 1.Forsimplicity,eachMAG inthenetwork as shown in Table 1.Table 1 shows the new setup val- topology is considered to have same service rate μ and ues needed by LB-CPMIPv6 mechanism for performing a threshold value θ . The simulation is conducted under load balancing. In addition, the performance gain is calcu- three different scenarios. In the first scenario, the LB- lated as in [31] to show the variation results between the CPMIPv6 mechanism is applied in the PMIPv6 network proposed LB-CPMIPv6 and LBM-PMIPv6 mechanisms (without-clustering) and is compared with LBM-PMIPv6 based on Eq. (13), where x and x ´ represent the results pro- mechanism [13]. This scenario can show the impact of LB- duced using LB-CPMIPv6 and LBM-PMIPv6 mechanisms CPMIPv6 mechanism in the PMIPv6 protocol that uses respectively. no clustering technique. In the second scenario, each MN can connect to one, two or three MAGs with an equal (x − x ´) probability within the CSPMIPv6 domain. This scenario Performance gain =  × 100 (13) (x ´) is performed to inject the overlapped area with a high number of MNs in order to show the actual impact of the intra-cluster handoff process on the system performance. 6.1 System setup In the last scenario, each MN is connected to one or zero This section illustrates the simulation setup used in the MAG with a probability equal and 0.2 and 0.8 respec- experiments in order to evaluate the proposed load bal- tively, without any concentration on the overlapped area ancing mechanism. The experiments are performed using between the MAGs. Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 17 of 23 LMA CN MAG MAG MAG MAG HMAG2 HMAG1 MAG MAG MAG MAG MAG MAG Fig. 11 The network topology used for simulation 6.2 Results and discussion from HMN. In this case, the LB-CPMIPv6 mechanism In this section, the performance metrics evaluated are works precisely similar to LBM-PMIPv6 mechanism, and the average queuing delay, handover latency, transmission PMIPv6 protocol shows the highest average queuing delay since it does not support the load balancing. Thus, the rate and the packet loss. The queuing delay is defined as the summation of the waiting time of each packet in packet service time inside the overloaded MAG buffer the queue per MAGs. The transmission rate is measured increases as a result of traffic from connected MNs. by calculating the average amount of data transmission In the second scenario, Fig. 13 illustrates the average from the MAGs for the entire simulation. Three scenarios queuing delay per packet at the MAG versus the overall are conducted to demonstrate the enhancement of LB- system load for the entire simulation time within a clus- CPMIPv6 mechanism in the clustered protocol against tered environment. In this scenario, the CSPMIPv6 proto- the LBM-PMIPv6 mechanism in addition to the scenario col is employed as an experimental environment to show situation of no load balancing. The scenarios have been the effectiveness of the proposed LB-CPMIPv6 mecha- conducted as follows. nism in the clustered domain, which totally differs from In the first scenario, the proposed mechanism applied the first scenario applied in the non-clustered domain. on the standard PMIPv6 protocol, which means that It is evident that the LB-CPMIPv6 and LBM-PMIPv6 the MAGs belong to the same domain (no hierarchical mechanisms outperform the average queuing delay of the domain). In this scenario, a comparison between LB- CSPMIPv6 protocol, which has no load balancing. Per- CPMIPv6 mechanism and LBM-PMIPv6 mechanism in forming the load balancing action leads to the reduction PMIPv6 domain is carried out in terms of measuring of the packets buffering time at the overloaded MAGs. average queuing delay. The result of this comparison is Furthermore, the performance for LB-CSPMIPv6, LBM- depicted in Fig. 12. As observed from the figure both, PMIPv6 and CSPMIPv6 shows same results regarding the the LB-CPMIPv6 mechanism and LBM-PMIPv6 mech- queuing delay before the predetermined thresholds take anism have achieved similar results in terms of average place. queuing delay, while PMIPv6 protocol, which has no load When the system load reaches 0.105 and 0.175, the balancing mechanism, shows a higher average queuing HMAG and/or the LMA starts to perform a load balanc- delay. In the LB-CPMIPv6 mechanism, if there is only ing action because the values in the figure are influenced one domain (no clustering), the serving MAG selects the after the predefined thresholds take place. In addition, TMAG according to its load and based on RSS reported when the total load reaches 0.35 for θ = 0.5, a few f Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 18 of 23 LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 12 The average queuing delay per MAG in the PMIPv6 protocol MAGs suffer from the overload issue. Subsequently, the affects the intra-cluster handoff. The performance gain of overloaded MAGs start performing a load balancing pro- LB-CPMIPv6 mechanism over LBM-PMIPv6 mechanism cedure as depicted in Fig. 13. The LB-CPMIPv6 mecha- is 15.66% on average. nism shows identical results regarding the queuing delay The LB-CPMIPv6 mechanism selects the ideal MAG in comparison to the LBM-PMIPv6 mechanism before from the best candidates through employing the domain reaching the respective thresholds. On the other hand, number, which in turn leads to shortening the time when the respective thresholds are reached, the average needed for delivering packets to their destination after queuing delay increases whenever the load increases. This the handoff. For example, when the load reaches 0.105 isduetotheincrease of thenumberofexchanged pack- on the HMAG, the HMAG performs load balancing ets among the MNs. However, the LBM-PMIPv6 mecha- by sending a message to the overloaded MAG, which nism shows high queuing delay compared to LB-CPMIPv6 is located on its domain and each overloaded MAG mechanism, as illustrated in Fig. 13.Thisisbecause sub- selects the TMAG based on its load, domain number and domains are not utilized in the clustered protocols, which the RSS. LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 13 The average queuing delay obtained from scenario 2 Avg. queuing delay (ms) Avg. queuing delay (ms) Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 19 of 23 Figure 14 illustrates the packet loss ratio between LB- different domain, which increases the buffering time. This CPMIPv6, LBM-PMIPv6 mechanisms and the CSPMIPv6 results in the increasing of congestion from the queuing protocol. The advantages of applying the load balancing system perspective. mechanisms in the protocols can be demonstrated clearly A related example to this is the movement of the HMN to an overall reduction in the overloaded MAGs. For to another cluster domain causes an extra delay due to example, the load balancing reduces the increased level the time needed to exchange its information among the of buffer utilization, which in turn reduces the number HMAGs. After this, this information should be emulated of lost packets. In addition, the LB-CPMIPv6 mechanism to the TMAG, which in turn needs a buffering technique achieves better results in terms of packet loss as com- to preserve the packets during the handover process. pared to the LBM-PMIPv6 mechanism. This is because Figure 15 compares the effects of LB-CPMIPv6 mech- the fact that the LB-CPMIPv6 mechanism moves the anism and LBM-PMIPv6 mechanism on the handover HMNs within the same domain as possible, which results latency. The handover latency is the interval between the to bring down the time of the handoff process leading to time of the last packet that is received by the HMN from the reduction packet loss. In other words, shortening the the old path and the time of the first packet that is received time needed to perform the handoff process leads to the from the new path by the HMN. The LB-CPMIPv6 reduction of the packet waiting in the buffer, which in turn mechanism outperforms the LBM-PMIPv6 mechanism in reduces the packet loss. terms of the handover latency. This is due to that the When the total load reaches 0.175, the LMA and a min- TMAG is selected based on the domain factor. In other imum one of the HMAGs exceed their thresholds, which words, the time needed to perform a handoff process are depicted in Table 1 (θ and θ), and for that the bal- by the HMNs is reduced. This is done by eliminating ancing function will be triggered by LMA and/or the the authentication process on TMAG and performing the respective HMAG. The TMAG that has taken the same handover without the involvement of the LMA, which domain with the MAG that has load balancing action will may be located far from the HMNs. The handover pro- be selected according to the MAG selection criteria in cess is performed by the overloaded MAG if the load in LB-CPMIPv6 mechanism. This definitely leads to short- the LMA, HMAG or MAG exceeds their predetermined ening the time needed to register the HMN on the TMAG, thresholds. As in Fig. 15, the handover is started, when the which in turn decreases the packet waiting time in the total load reaches 17.5 and is performed again when some MAGs become overloaded. The performance gain of LB- queue. This enhancement leads to the reduction of the packet congestion from the point of view of the queu- CPMIPv6 mechanism over LBM-PMIPv6 mechanism is ing system. Moreover, selecting the TMAG based on the almost 32.68%. RSS and load status only in the LBM-PMIPv6 mechanism In the third scenario, the average queuing delay, trans- increases the packet loss ratio. This is due to the long reg- mission rate and end-to-end delay are measured, as shown istration time that is needed to register the HMN in a in Figs. 16, 17,and 18 respectively. LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 14 Number of packet loss for LB-CPMIPv6 mechanism, Kim and Lee [13] and non-load balancing Packet Loss Ratio (%) Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 20 of 23 LB-CPMIPv6 mechanism LBM-PMIPv6 [13] 4.5 3.5 2.5 1.5 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 15 Impact of intra-cluster handoff on the handover latency Figure 16 depicts the impact of the LB-CPMIPv6 on The performance gain of LB-CPMIPv6 mechanism over the average queuing delay in comparison with the LBM- LBM-PMIPv6 mechanism is almost 9%. PMIPv6 mechanism and the original CSPMIPv6 proto- Figure 17 shows the average data transmission rate from col. The figure shows that the LB-CPMIPv6 mechanism the MAGs per MNs in the third scenario. The MNs increases the performance of the LBM-PMIPv6 mecha- scattered randomly within the CSPMIPv6 domain. It is nism even when the overlapped area is characterized by a obvious that the LB-CPMIPv6 mechanism has a higher small number of MNs. This can be attributed to the pro- data transmission rate than the other mechanisms, while posed mechanism performance to select the TMAG from the CSPMIPv6 with no-load balancing has the least data the same domain when the overloaded MAG performs a transmission rate. load balancing action. This definitely leads to the short- We observed from Fig. 17 that the data transmission rate ening of the time needed to register and authenticate the is roughly stable in the LB-CPMIPv6 and LBM-PMIPv6 MN on the TMAG, which in turn decreases the packet mechanisms. However, in the case of no-load balancing, waiting time in the queue, especially in the limit queue. the data transmission rate decreases whenever the MNs LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 16 The average queuing delay obtained from the third scenario Avg. queuing delay (ms) Avg. Handover Latency (ms) Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 21 of 23 0.9 0.8 0.7 0.6 0.5 0.4 LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0.3 0 100 200 300 400 500 600 MN’s arrival rate to the MAGs Fig. 17 The transmission rate obtained from the third scenario arriving rate increases. This is due to the absence of load the HMN and authenticating it in the TMAG, which in balancing that leads to an unbalanced situation at the turn increases the amount of sent packets to their targets. MAGs within the CSPMIPv6 domain. Furthermore, LB- However, in the LBM-PMIPv6 mechanism, the traffic usu- CPMIPv6 mechanism shows a significant enhancement in ally has an extra delay as a result of sending packets to data transmission rate compared to LBM-PMIPv6 mech- another cluster. anism, as shown in Fig. 17. This is due to the fact that Figure 18 presents the measured average of the end- the LB-CPMIPv6 mechanism gives higher priority for to-end delay per MAG in the CSPMIPv6 versus the total the selection of the TMAG based on its domain without load on the overall system. Interestingly, the LB-CPMIPv6 affecting the load status or the SS threshold. This mech- mechanism outperforms the LBM-PMIPv6 mechanism anism increases the traffic among the MAGs resulting in and the CSPMIPv6, which has no load balancing, despite increasing the data transmission rate. Furthermore, for- reducing the overlapped area. This is due to performing warding the HMNs traffic to the TMAG that is located in load balancing action by the overloaded HMAGs, which the same cluster reduces the time needed for registering leads to distribute the load within their clusters. This, LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 18 End-to-end delay per MAG versus the total load Avg. End-To-End delay (ms) Avg. Dtata Rate/MN from MAGs (Mbps) Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 22 of 23 in turn, leads to shortening the routing path as well as Acknowledgments Thanks to all my family, Sons, friends, and my colleagues who helped me to choosing the most closest MAG to the serving MAG. achieve this work. A special thanks to my supervisor Dato Prof. Shamala Selecting the TMAG based on its domain number has Subermaniam. a positive impact on the overall system performance. Availability of data and materials Subsequently, the result shows that the utilizing domain NS2 simulator and PMIPv6 patch are employed to test the work in this paper. number during the TMAG selection reduces the end- to-end delay due to the path faster recovery for the Authors’ contributions SMG designed the methods, conducted the experiments, evaluated HMN after the handoff. Furthermore, the original CSP- performance, and wrote the paper. SS defined the research area, problems MIPv6 protocol has a higher end-to-end delay due to and objectives. ZAZ and AM have equally contributed by given the final the unfair distribution of load. The unbalancing MAGs approval of the version to be published and contributed to the analysis. All authors read and approved the final manuscript. suffer from the heavy load, which in turn increases the overhead on the MAGs queue, causes extra packet time Authors’ information delay. Moreover, CSPMIPv6 tends to have relatively Safwan Ghaleb has received his bachelor degree in Computer Science from University of Jordan, Amman, Jordan, in 2009, the master degree in computer long paths, which also contributes to increasing the science from Jordan University of Science and Technology, Irbid, Jordan in end-to-end delay. 2012. He is working towards Ph.D. in computer networks, Universiti Putra Malaysia. His research interest include Internet of Things (IoT), Wireless and Mobile Networks, and Data Mining. 7Conclusions S. Shamala received the B.S. degree in Computer Science from University Putra PMIPv6 protocol and its extensions have been proposed Malaysia (UPM), in 1996, M.S. (UPM), in 1999, Ph.D. (UPM) in 2002. Her research to provide a seamless handover action within a localized interests are computer networks, simulation and modeling, scheduling and real time system. Dr. Shamala is now Prof. at the Department of management network. This is achieved via relieving the Communication Technology and Networks, Faculty of Computer Science and MN from any signaling-related to the mobility process Information Technology, University Putra Malaysia (UPM), Malaysia. when the MN changes its link. This is done by adding the Dr. Zuriati Ahmad Zukarnain is a professor at the Faculty of Computer Science and information Technology, University Putra Malaysia. She has served as a new MAG that performs the mobility related-signaling Head of Department of Communication Technology and Networks at the with the LMA instead of the MN. Furthermore, the MAG Faculty of Computer Science and information Technology, University Putra establishes a tunnel with LMA to send and receive the Malaysia. She received her PhD from the University of Bradford, UK. Her research interests include: Efficient multiparty QKD protocol for classical packets of the MN. However, to establish a new link con- network and cloud, load balancing in the wireless ad hoc network, quantum nection, the MN has to be associated with a specific processor unit for quantum computer, Authentication Time of IEEE 802.15.4 MAG. This association could overload the MAG. Conse- with Multiple-key Protocol, Intra-domain Mobility Handling Scheme for Wireless Networks, Efficiency and Fairness for new AIMD Algorithms and A quently, the LB-CPMIPv6 mechanism has been proposed Kernel model to improve the computation speedup and workload in this article to fairly distribute the loads among the performance. She has been actively involved as a member of the editorial MAGs fairly. The main advantage of LB-CPMIPv6 is its board for some international peer-reviewed and cited journals. Dr. Zuriati is currently undertaking some national funded projects on QKD protocol for capacity to consider clustered domain within the clustered cloud environment as well as routing and load balancing in the wireless ad protocols, which is not considered in other competitive hoc network. Dr. Zuriati is the founder of ZA Quantum Sdn Bhd, the start up mechanisms. company from University Putra Malaysia to produce a software designing tool for Quantum Communication known as Quantum Communication Simulator In the LB-CPMIPv6, the HMN that has a real-time ses- (QuCS). sion will not be selected during the process of the load Abdullah Muhammed received the bachelor degree in computer science from balancing; this restriction relieves the critical applications Universiti Putra Malaysia in 1998, the master degree in computer science from Universiti Malaya in 2004 and the PhD degree in computer science from the from service disruption. Furthermore, the CSPMIPv6 University of Nottingham, United Kingdom in 2014. He is a senior lecturer at handover signaling has been extended to be adapted with the Department of Communication Technology and Networks, Faculty of the newly proposed load balancing mechanism. Moreover, Computer Science and Information Technology, Universiti Putra Malaysia and is currently the HoD. His main research interests include cloud computing, the LPBA, PBA, and the heartbeat messages are modified mobile and wireless network, scheduling, heuristic and optimization. to enable sharing of the domain number for the new load balancing mechanism. Competing interests The authors declare that they have no competing interests. The LB-CPMIPv6 mechanism is implemented and simulated using the well-known NS2 simulator. The Publisher’s Note evaluation of the LB-CPMIPv6 mechanism in com- Springer Nature remains neutral with regard to jurisdictional claims in parison to the LBM-PMIPv6 load mechanism and published maps and institutional affiliations. CSPMIPv6 protocol is performed in terms of queu- Author details ing delay, packet loss ratio, end-to-end delay, and Department of Communication Technology and Network, Universiti Putra transmission rate. The results show that the new 2 Malaysia, 43400 Serdang, Selengor D.E., Malaysia. Sports Academy, Universiti load balancing mechanism achieves a better perfor- Putra Malaysia., 43400 Serdang, Selengor D.E., Malaysia. mance by reducing the average queuing delay, packet loss, end-to-end delay, and increasing the transmission Received: 20 October 2017 Accepted: 27 April 2018 rate. Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 23 of 23 References Proxy Mobile IPv6 networks: an IP multicast perspective (IEEE, Honolulu, 1. PCD Johnson, J Arkko, Mobility support in IPv6. Technical report, RFC 2014), pp. 766–770. https://doi.org/10.1109/ICCNC.2014.6785433 3775 (June 2004). https://www.rfc-editor.org/info/rfc3775 23. S Jeon, RL Aguiar, N Kang, Load-balancing Proxy Mobile IPv6 Networks 2. ELKDVCKS Gundavelli, B Patil, Proxy Mobile IPv6. Technical report, IETF, with mobility session redirection. IEEE Commun. Lett. 17(4), 808–811 RFC 5213 (2008). https://www.rfc-editor.org/info/rfc5213 (2013) 3. SM Ghaleb, S Subramaniam, ZA Zukarnain, A Muhammed, Mobility 24. C-M Huang, M-S Chiang, PB Chau, in 2015 IEEE International Black Sea management for IoT: a survey. EURASIP J. Wirel. Commun. Netw. 2016(1), Conference on Communications and Networking (BlackSeaCom).A 165 (2016) load-considered fast media independent handover control scheme for 4. AJ Jabir, SK Subramaniam, ZZ Ahmad, NAWA Hamid, A cluster-based Proxy Mobile IPv6 (LC-FMIH-PMIPv6) in the multiple-destination proxy mobile IPv6 for IP-WSNs. EURASIP J. Wirel. Commun. Netw. 2012(1), environment (IEEE, Constanta, 2015), pp. 171–175. https://doi.org/10. 1–17 (2012) 1109/BlackSeaCom.2015.7185109 5. JH Lee, KD Singh, JM Bonnin, S Pack, Mobile data offloading: a host-based 25. SM Raza, D Park, Y Park, K Lee, H Choo, in Proceedings of the 10th distributed mobility management approach. IEEE Internet Comput. 18(1), International Conference on Ubiquitous Information Management and 20–29 (2014) Communication. IMCOM ’16. Dynamic load balancing of local mobility 6. MM Islam, E-N Huh, Sensor proxy mobile IPv6 (SPMIPv6)—a novel anchors in software defined networking based Proxy Mobile IPv6 (ACM, scheme for mobility supported IP-WSNs. Sensors. 11(2), 1865–1887 (2011) New York, 2016), p. 106. http://doi.acm.org/10.1145/2857546.2857654 7. MM Islam, S-H Na, S-J Lee, E-N Huh, A novel scheme for PMIPv6 based 26. Y Kyung, Y Kim, K Hong, H Choi, M Joo, J Park, in 2016 IEEE Symposium on Wireless Sensor Network. Future Generation Information Technology: Second Computers and Communication (ISCC). Mobility-aware load distribution International Conference, FGIT 2010, Jeju Island, Korea, December 13-15, scheme for scalable SDN-based mobile networks (IEEE, Messina, 2016), 2010. Proceedings. (Springer, Berlin, 2010), pp. 429–438 pp. 119–124. https://doi.org/10.1109/ISCC.2016.7543725 8. MM Islam, TD Nguyen, AA Al Saffar, S-H Na, E-N Huh, in Computational 27. K Papagiannaki, N Taft, ZL Zhang, C Diot, in INFOCOM 2003. Twenty-Second Collective Intelligence. Technologies and Applications: Second International Annual Joint Conference of the IEEE Computer and Communications. IEEE Conference, ICCCI 2010, Kaohsiung, Taiwan, November 10-12, 2010. Societies. Long-term forecasting of Internet backbone traffic: observations Proceedings, Part III, ed. by J-S Pan, S-M Chen, and NT Nguyen. Energy and initial models, vol. 2 (IEEE, San Francisco, 2003), pp. 1178–11882. efficient framework for mobility supported smart IP-WSN (Springer, Berlin, https://doi.org/10.1109/INFCOM.2003.1208954 Heidelberg, 2010), pp. 282–291. http://dx.doi.org/10.1007/978-3-642- 28. EKRELHKNKSV Devarapalli, J Laganier, Heartbeat mechanism for proxy 16696-9_31 mobile IPv6. Technical report, RFC 5847, IETF Network Working Group 9. H-N Nguyen, C Bonnet, in 2008 5th IEEE International Conference on Mobile (June 2010). https://www.rfc-editor.org/info/rfc5847 Ad Hoc and Sensor Systems. Proxy mobile ipv6 for cluster based 29. R Jain, The art of computer systems performance analysis: techniques for heterogeneous wireless mesh networks (IEEE, Atlanta, 2008). https://doi. experimental design, measurement, simulation, and modeling. (Wiley, New org/10.1109/MAHSS.2008.4660097 York, 1991) 10. AJ Jabir, S Shamala, Z Zuriati, N Hamid, A comprehensive survey of the 30. S Deering, R Hinden, Internet Protocol, Version 6 (IPv6) Specification. current trends and extensions for the Proxy Mobile IPv6 Protocol. IEEE Technical report, IETF, RFC 2460 (1998). https://www.rfc-editor.org/info/ Syst. J. PP(99), 1–17 (2015) rfc2460 11. CKKRPBH Yokota, F Xia, Fast Handovers for Proxy Mobile IPv6. Technical 31. M Ghaleb, S Subramaniam, M Othman, Z Zukarnain, Predetermined path report, RFC 5949 (2010). https://www.rfc-editor.org/info/rfc5949 of mobile data gathering in wireless sensor networks based on network 12. MS Kim, S Lee, D Cypher, N Golmie, in Global Telecommunications layout. EURASIP J. Wirel. Commun. Netw. 2014(1), 51 (2014) Conference (GLOBECOM 2010). Fast Handover Latency Analysis in Proxy 32. T Issariyakul, E Hossain, Introduction to Network Simulator 2 (NS2). Mobile IPv6 (IEEE, 2010), pp. 1–5 (Springer, Boston, 2012), pp. 21–40 13. M-S Kim, S Lee, Load balancing and its performance evaluation for layer 3 33. K Fall, K Varadhan, The ns manual (formerly ns notes and documentation). and IEEE 802.21 frameworks in PMIPv6-based wireless networks. Wirel. The VINT project. 47, 19–231 (2005) Commun. Mob. Comput. 10(11), 1431–1443 (2010) 14. K Mun-Suk, L SuKyoung, A novel load balancing scheme for PMIPv6-based wireless networks. {AEU} - Int. J. Electron. Commun. 64(6), 579–583 (2010) 15. MS Kim, S Lee, in 2009 IEEE 20th International Symposium on Personal, Indoor and Mobile Radio Communications. Load balancing based on layer 3 and ieee 802.21 frameworks in pmipv6 networks (IEEE, Tokyo, 2009), pp. 788–792. https://doi.org/10.1109/PIMRC.2009.5450177 16. H Kong, Y Jang, H Choo, in Computational Science and Its Applications (ICCSA), 2010 International Conference On. An efficient load balancing of Mobile Access Gateways in Proxy Mobile IPv6 Domains (IEEE, Fukuoka, 2010), pp. 289–292. https://doi.org/10.1109/ICCSA.2010.67 17. J Dimple, C Kailash, A load reduction and balancing scheme for MAG operating in PMIPv6 domain. (Springer, Berlin, Heidelberg, 2013), pp. 128–136. http://dx.doi.org/10.1007/978-3-642-35864-7_19 18. S Qutub, T Anjali, in Electro/Information Technology (EIT), 2012 IEEE International Conference On. Load sharing mechanism for Mobile Access Gateways in PMIPv6 (IEEE, Indianapolis, 2012). https://doi.org/10.1109/EIT. 2012.6220760 19. F Teraoka, T Arita, in 2011 Third International Conference on Ubiquitous and Future Networks (ICUFN). PNEMO: a network-based localized mobility management protocol for mobile networks (IEEE, Dalian, 2011), pp. 168–173. https://doi.org/10.1109/ICUFN.2011.5949156 20. H Jiang, Load sharing support for MAGs in Proxy Mobile IPv6. Technical report, IETF-Draft (expired) (December, 2011) 21. T-T Nguyen, C Bonnet, Considerations of IP multicast for load balancing in Proxy Mobile IPv6 networks. Comput. Netw. 72(Supplement C), 113–126 (2014) 22. TT Nguyen, C Bonnet, in 2014 International Conference on Computing, Networking and Communications (ICNC). Load balancing mechanism for http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png EURASIP Journal on Wireless Communications and Networking Springer Journals
Free
23 pages

Loading next page...
 
/lp/springer_journal/load-balancing-mechanism-for-clustered-pmipv6-protocol-ob1Z5trbzT
Publisher
Springer International Publishing
Copyright
Copyright © 2018 by The Author(s)
Subject
Engineering; Signal,Image and Speech Processing; Communications Engineering, Networks; Information Systems Applications (incl.Internet)
eISSN
1687-1499
D.O.I.
10.1186/s13638-018-1137-y
Publisher site
See Article on Publisher Site

Abstract

Proxy Mobile IPv6 (PMIPv6) has become a credible member of pertinent research areas. This is attributed mainly to its capability of enabling mobility without imposing constraints or requirements on the mobile node (MN). This MN shield is enabled due to the transferring of mobility-related signaling to a new entity, which is called Mobile Access Gateway (MAG). However, associating MNs to a specific MAG inside the PMIPv6 network increases the MAG load probability. Thus, several research have enhanced the PMIPv6 protocol to improve its basic specifications and performance. Strategies include protocols, which apply the clustering technique to enhance the overall performance of the PMIPv6 in terms of routing, scalability, lifetime, and load balancing. The load balancing mechanism is considered in the non-clustered protocols. However, this mechanism has not been adopted in clustering-based protocols. Thus, pertaining to the load and the respective assignments is critical. In this article, to address these issues, a new load balancing mechanism is proposed among MAGs for Cluster-based Proxy Mobile IPv6 (CSPMIPv6) protocol. The signaling within the CSPMIPv6 has been enhanced to support the proposed load balancing mechanism. The proposed mechanism employs the inter- and intra-domain on a frequent basis to select the best MAG among the candidate MAGs. The new mechanism has improved the performance to create an evident improvement in terms of average queuing delay, handover latencies, transmission rate, end-to-end delay, and packet loss as compared to the LBM-PMIPv6 mechanism and CSPMIPv6 protocol. Keywords: Load balancing, Proxy mobile IPv6, IPv6 protocol, Queuing delay 1 Introduction the authentication, authorization, and accounting (AAA) In Mobile Internet Protocol (MIP), the high involve- server to register the MN with the respective LMA. The ment of MNs in the mobility-related signaling causes main role of LMA in the PMIPv6 protocol is to maintain several serious issues. Among the issues are long handover the MN accessibility whenever the MN changes its points latency and excessive signaling [1]. The MN is required of attachment within the PMIPv6 network. This removal to register with the home agent (HA) whenever the MN of responsibility from the MN results in the PMIPv6 changes its point of attachment. Addressing these prob- protocol enhancing the performance of MIPv6 protocol, lems associated with the MIP protocol, the Proxy Mobile especially in terms of traffic signaling, service disruption, IPv6 (PMIPv6) has been developed by the Internet Engi- and tunneling overhead. Therefore, making PMIPv6 a sig- neering Task Force (IETF) in order to effectively handoff nificant mobility management protocol for wireless sensor operations to MNs [2]. This is done by adding a new networks (WSNs). However, ignoring the load balancing entity, named Mobile Access Gateway (MAG) that takes among the MAGS and using single LMA to process or over the responsibility of mobility configuration from the forward the MN’s packets withing the LMA domain, have MN. The main role of the MAG entity is to detect MN resulted in many drawbacks (e.g, single point of failure, movement within the Local Mobility Anchor (LMA). In long handover latencies and intense signaling [3–5]). addition, the MAG initiates the required signals with To tackle these issues, research findings such as Sen- sor Proxy MIPv6 (SPMIPv6) [6–8], Cluster-based PMIPv6 *Correspondence: safwan_ghaleb@yahoo.com Department of Communication Technology and Network, Universiti Putra Malaysia, 43400 Serdang, Selengor D.E., Malaysia Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 2 of 23 for wireless mesh networks [9], andaCluster-basedProxy queuing delay. In the initial registration process of the Mobile IPv6 (CSPMIPv6) [4] have been developed to mit- MAGs and HMAGs, LB-CSPMIPv6 enables the LMA igate these problems. All these protocols have employed to assign a number for every sub-local domain in the clustering strategy in order to be more efficient for mobile clustered PMIPv6 domain. This domain is carried out users. The CSPMIPv6 [4]protocolsolvedahighnumber by the heartbeat message along with the load sta- tus in order to select the best MAG for the handoff of issues associated with the PMIPv6 and SPMIPv6 pro- tocols. Thus, the protocol is able to be used in a variety MN with the same domain of the MN’s serving MAG, of applications more compared to other protocols [3, 10]. which is different from existing schemes where each The CSPMIPv6 has inherited other drawbacks due to TMAG is selected based on its domain and load. In the dependency on the central and single LMA. The fast han- handoff process, LB-CPMIPv6 comprehensively consid- dovers for Proxy MIPv6 (PFMIPv6) [11]protocolhasbeen ers the scenarios of intra- and inter-handoff mobility to developed by the IETF to reduce the handover latency. provide a seamless mobility support to MHs roaming However, the serving network causes false handover initi- across various access networks, and low buffering cost, ation, due to the prediction of the target network to which which reduces handoff delay and prevents packet loss. theMNwillmove[12]. In this work, the CSPMIPv6 protocol handover signal- Contradictory to the benefits of the PMIPv6 proto- ing forms the core of the newly proposed load balancing col and its extensions, the constraints are caused by mechanism. The performance analysis of the proposed the MNs, which have to connect to a particular MAG load balancing mechanism with an extensive simulation within the PMIPv6 network. This causes the MAG to has been developed using Network Simulator (NS2) to be overloaded, especially in large networks. The over- show that the proposed load balancing mechanism (LB- loaded MAG causes a queuing delay, which in turn leads CSPMIPv6) achieves an improved quality of service (QoS) performance degradation packet loss, end-to-end delay, demands. and the throughput. There has been no consideration of In this work, the unique adoption of a load balancing load balancing in the basic specification of the PMIPv6 mechanism is developed to improve the overall system and its extensions. Thus, many types of research such performance of clustered PMIPv6 domain. as [13–18]haveattempted to solvethisissuethrough The main contributions of this article are as follows: applying the load sharing mechanism between the MAGs. 1. A detailed analysis of the CSPMIPv6 protocol in This has seen the increasing performance of the overall terms of merits, demerits, and its architecture, which system. Their proposed mechanisms, which are elabo- represents the underlying of the LB-CPMIPv6 rated in Section 3, deployed load balancing action by mechanism, is presented. The benchmark that has selecting the best target MAG, in addition to select- been selected for comparison purpose is reviewed ing the low-priority traffic MNs for the handoff process. extensively. These protocols have achieved good results in terms of 2. Providing an extensive overview of proposed striking a balance of load between the MAGs. How- mechanisms within the PMIPv6 domain. ever, these proposed mechanisms are applied only to 3. The development of a new load balancing non-clustered protocols. The clustering-based protocols mechanism for clustered PMIPv6 enhances the load have not researched, despite their widely being used in distribution among the MAGs within the CSPMIPv6 the research areas. Several issues such as high queu- domain. The focal point in this new mechanism is ing delay, end-to-end delay, and packet loss are accused exploiting the clustering benefits inside the PMIPv6 through applying these mechanisms when no consider- domain to enhance the process of selecting the ation is given to the division of clusters. Subsequently, TMAG during the handoff action. this is leading to serious disruption. As a result, it is evi- dent that the MAG selection has enormous potential for This article is organized as follows: enhancement, which is the focus of this article. The abil- Section 2 presents an extensive review of the CSPMIPv6 ity for serving network to select the Target MAG (TMAG) protocol focusing on its advantages, disadvantage, and in according to its domain will definitely lead to the reduc- particular the handover signaling. Section 3 deliberates in tion of the handover latency, end-to-end delay, and the detail the related work on the loading balancing in the average queuing delay. This is the result of the reduction PMIPv6 protocol. Section 4 discusses in detail the pro- of the signaling registration and the avoidance of the LMA posed LB-CPMIPv6 mechanism. In Section 5, a detailed involvement. explanation of the load balancing signaling for the clus- In order to achieve these potentials to increase the per- tered PMIPv6 domain is done and followed by Section 6, formance of the system, a load balancing based on the where the system architecture that is used as the environ- clustered PMIPv6 protocol is proposed LB-CSPMIPv6 to ment for the LB-CPMIPv6 mechanism is presented and provide a seamless mobility management and lowering the performance evaluation for LB-CPMIPv6 mechanism Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 3 of 23 is discussed. Section 7 concludes the contributions of the from the LMA, in order to mitigate the load and the proposed work. signaling on the LMA. In addition, the (AAA) func- tionalities are provided by the HMAG to reduce the 2 An overview of the CSPMIPv6 protocol registration time that is needed to register the MN. In this section, an extensive description of the CSPMIPV6 The registration processes of the new MN in the CSP- protocol, which has been used as a basis for the LB- MIPv6 protocol are performed according to the following steps: CPMIPv6 mechanism, is presented. Jabir et al. [4] proposes the clustered PMIPv6 archi- tecturetoovercomeproblemsassociatedwiththeProxy 1. Once the movement of MN has been detected by MIPv6 (SPMIPv6) [6] and Proxy Mobile IPv6 (PMIPv6) MAG , it sends a request message authentication to [2] protocols respectively. In this developed solution, the the AAA server including the MN identifier (MN-ID). PMIPv6 domain was divided into local sub-domains, 2. Then the MAG registers the MN in its domain as shown in Fig. 1. Each sub-domain contains several cluster by sending a Local Proxy Binding Update MAG clusters and each cluster is controlled and man- (LPBU) to the HMAG . aged by a cluster Head MAG (HMAG). As deliber- Upon the successful authentication by the HMAG , ated in the earlier sections, the CSPMIPv6 is derived the HMAG registers the MN on the LMA by sending from the PMIPv6, so functionalities of entities such as a Proxy Binding Update (PBU) including the MN-ID LMA, MAG, MN, and corresponding node (CN) are and the HMAG-ID. identical to those in PMIPv6 protocol. The new entity 3. Once the PBU message is received successfully by the HMAG in the CSPMIPv6 protocol has been configured LMA, a new Binding Cash Entry (BCE) is created to to take the responsibility of the local cluster handoff store the MN-ID and HMAG identifier. Internet CN LMA HMAG1 HMAG2 MAG1 MAG2 MAG3 MAG4 Intra-cluster Handoff Inter-cluster Handoff Fig. 1 Overall CSPMIPv6 system architecture Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 4 of 23 Subsequently, the LMA sends a Proxy Binding domain of the serving MAG. Thus, when the destination Acknowledgment (PBA) reply to the HMAG .The MAGacceptstoregisterthe MN,anLPBUmessage issent PBA message includes the Home Network Prefix to the respective HMAG including the MN-ID. However, (HNP) of the MN, which hereafter will be used for the HMAG will not find the MN-ID in its binding table maintaining the MN reachability within the PMIPv6 as the MN comes from another cluster. Therefore, the LMA must be involved in the process. This is done by the domain. The LMA configures the routing path with the HMAG by setting a bi-directional tunnel between requesting HMAG sending a PBU message to the LMA them to send and receive the traffic. advertising the new location of the MN. Subsequently, the LMA will update its BCE tables and send a reply to the 4. The HMAG adds the MN information to its Binding requesting HMAG. Update List (BUL) in order to register the MN and Once the requesting HMAG receives the PBA, a new sends a Local Proxy Binding Acknowledgment binding table for the MN will be created and a reply will (LPBA) message to the MAG containing the MN also be sent to the respective MAG. Finally, a new entry prefix. Then, routing configuration is performed to for the MN will be created by the MAG in its binding table make the MN accessible. and an HNP message will be sent to the MN. 5. When MAG gets the LPBA message from the The CSPMIPv6 has gained several substantial benefits HMAG , its BUL will be modified by adding the MN as a result of dividing the PMIPv6 domain into sub-local and forward the HNP to the MN through the networks. These advantages have increased the MN user advertisement message. Now, the MN has the ability performance concerning the mobility management. This to send and receive traffic. performance enhancement comes as a result of reduc- The MN information at the end of registration will be ing the LMA load by relieving it from the local mobility stored in the MAG ,HMAG ,andLMAtablesasstatedin i j signaling within the HMAG cluster. Furthermore, the sig- the aforementioned registration operation. Furthermore, naling cost has been reduced as a result of integrating The HMAG exchanges the MN information with the the AAA functionalities with the HMAG. Another critical MAG to perform a routing configuration for the MN. benefit is shortening the routing path when the MN moves Thus, there will be no need for a bi-directional tunnel inside its cluster (i.e., intra-cluster handoff) while per- set up between the HMAG and MAG [19]. Moreover, j i forming the handoff process by the HMAG without any the idea of integrating the AAA functionalities with the involvement from the LMA. Despite all of these merits LMA functions proposed by [6]isreusedinthisCSP- mentioned above, the CSPMIPv6 still suffers from sev- MIPv6 protocol to reduce the signaling cost during the eral issues such as the one point of failure (single LMA), MN registration. end-to-end delay, and excessive signaling [3]. The handoff procedure within the CSPMIPv6 domain functions is illustrated in Fig. 2 and deliberated as follows. 3 Related works on PMIPv6 protocol load When the MN decides to move from its serving network balancing mechanisms to another within the CSPMIPv6 domain, the MN move- In the PMIPv6, the mobility-related signaling responsibil- ments could be either an intra- or inter-cluster handoff. ity is undertaken by the MAGs on behalf of the MN. All In the intra-cluster handoff, the MN is supposed to move the MNs must be connected to a particular MAG which to another MAG within the same cluster domain. In other makes the MAG overloaded easily. The overloading on the words, the MN movement is still controlled by the same MAG leads to an increase of packet loss, end-to-end delay, cluster head HMAG. Therefore, the handover here is per- and the decrease of the transmission rate. Consequently, formed by the HMAG through updating its binding table several works have been proposed to reduce the load on without any intervention from the LMA. To do so, the the overloaded MAGs via applying the load sharing mech- destination MAG to which the MN decides to move, will anism between the MAGs to avoid the negative effect on send an LPBU message to the respective HMAG includ- the overall system performance. ing the MN-ID. Here, the HMAG will only need to update Kim and Lee [14] propose a load-balancing mechanism its table by setting the new MAG address in its MAG field to equitably distribute the load among different MAGs as opposed to the inter-cluster handoff. This is done once within the PMIPv6 domain. The proposed work led to the MN information has already been recorded. Then, the improving the overall system performance in terms of respective HMAG sends back an LPBA message, includ- average queuing delay, packet loss, and end-to-end delay, ing the HNP to the requesting MAG as well as configures while increasing the transmission rate. The authors uti- the routing performed with the requesting MAG in order lized the heartbeat message in order to allow for a specific to forward the MN packets. MAG to learn the load status of its neighboring MAGs. In the inter-cluster handoff, the MN movement is The heartbeat message is modified in order to store the detected by another MAG located outside the cluster load field for the load balancing action. Similarly, the Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 5 of 23 S-MAG1 MAG3 LMA HMN MAG2 HMAG1 HM AG2 MN occurrence detected LPBU message BUL Updating LPBA message BUL Updating MN occurrence detected LPBU message PBU message PBA message Updates BUL LPBA message BUL Updates Router Advertisement (HNP) Fig. 2 Handoff procedure in the CSPMIPv6 domain MAGs also sends a heartbeat message to the LMA, includ- reused in this work to create an identical platform for ing their load status. The LMA stores the received loads comparative purpose. in its BCE used to measure the overall system perfor- Another work in this area has been done by Kim and mance. The description of this is shown in Fig. 3.When Lee [13, 15] to enhance the load balancing by utilizing the LMA load exceeds a specific threshold, a heartbeat the IEEE 802.21 standard. The IEEE 802.21 optimizes message is sent by the LMA to the overloaded MAG. the handover between the heterogeneous technologies via Then, the overloaded MAG performs a load balancing facilitating media-independent handover by providing up and chooses the MNs that have the option to change layers with network-related information. This work aims their point of attachment. The target MAG is selected by to determine the load on a candidate point of attachment the serving MAG based on the received signal strength (PoA). There are cases where the PoA suffers from heavy (RSS) and the load status reported from the MNs. The loads as compared to the TMAG that experiences a lower signalingprocessthatis performedduringtheload bal- amount of load. This happens if the MAG load concen- ancing action is presented in Fig. 4.Thisworkrestricts trates on only one of its PoA (BS/AP). Thus, the target the procedure of choosing the handover MN (HMN) for PoA load is very important to knowing to reduce the over- the handover process by preventing the serving MAG to all load overhead. This proposed technique has proven to select the MNs that have a real-time session. Numerical have a remarkable enhancement in terms of queuing delay and simulation analysis has been conducted by the authors and transmission rate. to evaluate their proposed mechanism, and their result Another load balancing approach has been proposed shows significantly enhanced performance over the orig- by Kong et al. [16] for efficient migration of the load inal PMIPv6. The abovementioned mechanism forms the between the MAGs. Their approach determines the tar- core of the proposed LB-CPMIPv6 mechanism. Further- get MAG which needs a low signaling requirement. Each more, all the paper variables and assumptions are also MAG learns the load status of its neighboring MAGs by Inter-handoff-cluster Intra-handoff-cluster BCE Updating Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 6 of 23 MAG LMA Start Start Set initial values for Set initial values for ϴs ϴf, μ s,Nm ϴ ϴf No No On Packets On Packets receiving Receiving Yes Yes No No p >ϴ i i pt >ϴ Yes Yes No f<ϴf MAG preforms a load balancing action Yes MAG select the handover MNs (HMN) LMA sends load balancing request message to MAG MAGi selects the best MAG from the non- i overloaded candidate MAGs MAG obtains the MAGs load from the Heartbeat message exchanges among the MAGs MAG sends handover command to HMNs to change their routing to the TMAG MAG sends load status to LMA periodically Fig. 3 Load balancing operation in the PMIPv6 domain [13] exchanging their load among each other in the domain. makes the MN moves to a different domain that requires Then, the MAGs create a list of candidate MAGs based extra signaling, which in turn leads to high queuing delay on the received load information in order to select the and low transmission rate. best TMAG for the HMN. A proactive load balancing is An agent-based scheme was proposed by Dimple and performed during the initial attachment of the MN by Kailash [17] to mitigate the overloaded MAG issue within selecting the MAG that has the lowest load according the PMIPv6 network. Their mechanism works by mov- to the load information. This is done before the current ing the mobile agent from one location to another to MAG becomes overloaded. Therefore, by avoiding the reduce the load on the overloaded MAG. The mobile overloaded MAG, benefits such as low packet loss and agent achieves this through visiting one MN to collect its low signaling will be achieved. However, in this mecha- data and moves to the other MNs associated to the MAG nism, the HMN experiences an extra delay, especially in to take the only relevant data for transmission, in order to the proactive scenario caused by the time needed by the reduce the overhead communication. The MN selection is serving MAG to determine the best TMAG to which the performed according to certain criteria such that the MNs that have real-time session will not be selected, while the HMN moves according to the MAGs loads. Real-time ses- sions have not been considered in [16], which in turn MNs that have high-rate data connection become a target degrades the system performance. Moreover, this mecha- for a handoff. Despite the benefits gained by employing nism requires MNs with multi-interface to be connected the MN agent, several issues arise. Anticipating the MN with two different networks, which makes it restricted in the load balancing adds some burden to the MNs and to this scenario. Also, multi-domains within the same increase the function complicity. This is done by selecting domain has not been considered in this work, which one MN to visit the other MNs within the MAG domain Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 7 of 23 HM N LMA SMAG C-MAG1 C-MAG2 Report (MN-ID, MN-IID, RSS, New MAG Info.) HI Message HI Message Determine the MAGs Load HAck Message (C-MAG1 load Info. HAck Message (C-MAG2 load Info. Determine the TMAGs for the HMN Handover Command(TMAG-Info.) Establish L2 connection by the MN MN Starts New Connection PBU Message PBA Message Router Solicitation (RS) Router Advertisement (RA) Fig. 4 Load balancing signaling in the PMIPv6 domain to collect the similar data packets which require some sig- that satisfy the mobile users. However, employing the naling messages between the MN and the associated MN. Global Position System (GPS) expedites the power of Moreover, the employed θ threshold by the LMA in the the MNs, which is not acceptable in the critical applica- mechanism depends on the size of the data reduction by tions . Besides, this work consecrates on the overloaded the MAG that sent to the LMA. This leads to overload MAGs and ignore the overloaded LMA. The overloaded the MAG that has numerous attached MNs but does not LMA is determined according to the all MAGs load have any similar data between them or have less than the in the system, which may be accrued even when the specified threshold, which not reflect the reality load state MAGs are not overloaded. This definitely degrades the of the MAG. Furthermore, clustered PMIPv6 protocol not overall system performance through increasing the time considered in their implementation, which in turn may of registering/de-registering the MNs (large queuing lead to effect the intra-domain mobility advantages in a delay). Furthermore, divided domains do not consider in contrary manner. their work, which affects the QoS regarding the mobility Qutub and Anjali [18] introduce an efficient mechanism management. The load balancing problem also has been researched by to balance the load among the overloaded and low-load MAGs. Their mechanism selects the target MAG accord- the Internet Engineering Task Force (IETF) and for which ing to its geographical serving area and its current load. a Request for Comments (RFC) was introduced by Jiang Also, the MN selection for handover is performed based [20]. Each MAG sends its load periodically to the LMA on the MN’s QoS profile, location, direction, and multi- and hereafter is used by the LMA to create a list of can- interface capability. This selection has proven to reduce didate MAGs for performing load balancing. The factors the overload and provide the service, which satisfies the have been used in their mechanism to select the target QoS. In this work, not only the overloaded is avoided MAG are specified in [18]. The process of selecting the but also the services are provided with a level of QoS HMN is performed as follows: Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 8 of 23 1. When the MAG becomes overloaded, the MAG these sessions according to same criteria as specified in starts load balancing action by selecting the HMN [21]. Their experiments and numerical results that have according to its service type to avoid selecting the been conducted have shown significant improvements in MN that has a real-time session. terms of load distribution as well as reducing the multicast service disruption. However, moving the MN’s multicast 2. Then, the MAG sends Load State Message (LSM) to session to another LMA (LMA has least load), which may the LMA in order to inform the LMA about its loads. 3. Accordingly, the LMA gives a feedback to the MAG be located far away from the MN affects the system per- about the overloaded MAGs and the non-overloaded formance. This is due to the long path between the MN MAGs. This is done by sending a Load State and the new LMA that leads to long delay or packets Acknowledgment Message (LSAM) to the MAG in drops high signaling cost. Moreover, this work focuses on order to migrate the HMN to a new MAG that is not balancing the load between the LMAs and ignores the overloaded. overloaded MAGs, which may overload even when the load on LMAs is balanced. 4. Once the MAG receives the LSAM message, it sends Another work focuses on distributing the load between a request message to non-overloaded MAGs. the LMAs introduced in [23]. The primary aim of this 5. The non-overloaded MAGs upon receiving the work is moving the load from the overloaded LMA to request messages reply to the requested MAG along the LMA has the least load. This is done when the load with the acceptance or the rejection of its request at LMA reached the specified threshold. Accordingly, the according to their status. LMA sends load balancing (LB) warning to the MAG that 6. Then, the overloaded MAG sends a notification to serves the selected MN. Then, the MAG sends refresh the HMN including the information about the binding to the LMA and hereafter the LMA communi- TMAG. cates with the new LMA to bind the selected MN to 7. The MN once receives the notification from the another LMA. Now, the MN anchored at the new LMA. MAG, it sends Router Solicitation (RS) message to This work shows remarkable improvements regarding the TMAG to inform it about its movements. the blocking probability and dropping probability than 8. A PBU and PBA messages are exchanged between PMIPv6 with no load control. However, the authors do the MAG and the LMA to register the handoff MN. not take into consideration the overloaded MAG, which 9. Finally, the TMAG sends a Router Advertisement is easily susceptible to be overloaded any time when the (RA) message to the HMN including the new IP attached MNs very high or when the MN requires a high addresses in order to complete its registration. stream session. This leads to service disruption through According to the registration procedures in this mech- increasing the queuing delay. In addition, the hierarchical anism, the MAG should send a request to the all non- domain is not considered also in their work, which may overloaded MAGs and await their responses to select the lead to high queuing delay. TMAG for the HMN based these responses. This process A load balancing scheme is introduced in [24]to consumes the bandwidth due to the messages exchanged improve the overall system performance in terms of between the MAGS during the load balancing mecha- handoff delay and throughput. The IEEE 802.21 Media nism. In addition, this work does not target the overloaded Independent Handover Services (MIH) functionalities LMA, which is responsible for the acceptability of all the are utilized with the proper selection of the MN new MNs connected to the MAGs that may be overloaded. network to provide a seamless handover in the heteroge- Moreover, the intra domain mobility in case of divid- neous networks. In this scheme, when the signal of the ing the PMIPv6 into sub-local domains is not considered, MN becomes very weak, a report from MN is sent to the which lead to long handoff delay through increasing the MN serving MAG. Then, the PMAG upon receiving the path recovery and signaling cost. report sends handover initiate (HI) message to the LMA Nguyen and Bonnet [21, 22] introduce a solution mech- including all candidates MAG/APs information. Accord- anism to solve the issue of load balancing in the PMIPv6 ingly, the LMA forwards the HI message to the candidates by considering the IP multicast session. Their solution and these candidates response to the LMA with sending a caters two scenarios, which are named as proactive- Handover Acknowledgment (HAck) messages to inform multicast and the reactive-multicast. For the former, when the LMA about their status and their acceptance to serve the MN. The LMA forwards the received HAck messages the MN starts a new multicast connection, a load balanc- ing action will be triggered to select the suitable LMA to the serving MAG in order to select the proper network to manage this connection. However, in the latter, when for the MN. Despite the enhancements that are made in the LMA becomes overloaded, the LMA starts to select this scheme regarding the handover time and throughput, some of the multicast sessions for a load balancing pur- additional signaling messages are required between the poses. Then, the LMA selects the suitable target LMA for PMAG, LMA, and the candidates MAGs/APs, which Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 9 of 23 negatively impact the system. The reactive load balancing issues focused on the clustered protocol and to provide is not considered in this scheme, which leads to increas- solutions. ing the blocking probability in the overloaded MAGs and In this work, the CSPMIPv6 protocol is implemented to service disruption due to increasing the queuing delay in make it as the central referral platform for the proposed the overloaded MAGs, in addition to ignoring the divided LB-CPMIPv6 mechanism. The proposed mechanism in [13] has been implemented in this research work whereas domain as the previous works do. Raza et al. [25] employ the Software Defined Network- it is applied on the CSPMIPv6 for a comparison pur- ing (SDN)-based solution in order to mitigate the loads pose. The CSPMIPv6 architecture is shown in Fig. 1.This between the LMAs. This works depend on central mobil- protocol divides the PMIPv6 domain into sub-domains. ity controller that is responsible for monitoring the load at Each domain encompasses some MAGs that form a clus- the LMAs. The controller upon detecting the load cross- ter within the PMIPv6 domain. Subsequently, each cluster ing over the predefined threshold on any of the LMA elects one MAG to act as the cluster head (HMAG). starts moving some traffic from the massive LMA load to The MAG in the CSPMIPv6 can be easily overloaded as the lower LMA load. According to the analytical analysis, in PMIPv6 protocol. Figure 5 shows an example of the their scheme has significant improvements regarding dis- CSPMIPv6-based inter-architecture of its overlapped area ruption period of uplink and downlink traffic during load among its clusters. The overlapped area between the sub- balancing action compared to their benchmark. How- domains contains a number of HMN candidates. These ever, adding extra element is costly. In addition, the mas- candidates must have another optional network to con- sive MAG load is not considered in their scheme, which nect with for the handover purpose. As seen in Fig. 5, affects the system performance in terms of handover MAG1 andMAG2are locatedinthe sameclusterasthe delay. Moreover, LMA domains also not consider in this HMAG1, while MAG3 is located in a different cluster scheme, which leads to moving the track to another LMA HMAG2. The solid lines represent the current connection located far away from the serving LMA. Furthermore, of the HMN candidate, while the dotted lines represent scalability issue has arisen as a result of using a central the optional connection for it. The selection process of controller. TMAG and HMNs must be performed to provide bet- SDN also used by [26] to reduce the blocking proba- ter performance to the HMNs. Likewise, choosing the appropriatenetwork fortheHMNsin theoverlapped bility and increase the resource utilization through using area leads to the balance of load between the MAGs, mobility-aware load distribution for multiple controllers. The objective of this work is handling the handover which in turn avoids the overlapped MAGs. The proposed messages as fast as possible. This is performed by dis- enhanced load balancing algorithm is presented in the tinguishing the handover messages (gives them high pri- next section. ority) and manipulate them by the controller has the least load among the other controllers if the serving 4 The proposed LB-CPMIPv6 mechanism controller suffers from heavy load. However, the main In this paper, a new mechanism, named LB-CPMIPv6 consideration is given to the loads on the LMA and is is proposed to enhance the overall system performance ignored the loads on the MAGs. In addition, the clus- of IP-WSNs by considering a load balancing approach tered domain also is not considered in their scheme. These in the clustered network. The proposed LB-CPMIPv6 ignoring lead to serious issues regarding the mobility, mechanism expands the MAG capability to avoid over- which in turn affect the service delivered to the mobile loading issue by developing a new load balancing mech- users. anism. In addition, the proposed mechanism reduces the The review of these deliberated algorithms raises some time needed to recover path between the communicating majorconcernswhichhavetobeconsideredfor theload entities. sharing mechanism. A list of candidate MAGs to be cre- In the proposed LB-CPMIPv6 mechanism, a domain ated with a fewer message exchange to avoid the network number should be assigned to every sub-domain in overloading issue and choosing the HMNs should be order to distinguish between the clusters within the performed based on their traffic type to avoid the selec- PMIPv6 domain. The proposed LB-CPMIPv6 mecha- tion of the HMNs that have an arguing critical-session. nism provides an efficient way to balance the load Unfortunately, proposed works above metioned have not between the MAGs, by predicting the proper TMAG proposed such solution for the clustered-based protocol to which HMN moves accurately, as illustrated in Algo- during the formation of the candidate MAG list. Thus, is rithms 1, 2, and 3. Algorithms 1, 2, and 3 explain effected the overall system performance as the selection the functionalities of MAG, HMAG, and LMA respec- of TMAG from another cluster or in the case, there is tively within the proposed mechanism. The control another target MAG from the same cluster of the serving flow diagram of LB-CSPMIPv6 mechanism is illustrated MAG. Thus, this work has been motivated by these open in Fig. 6. Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 10 of 23 MAG MAG MAG1 MAG3 HMN HMAG 2 MAG HMAG 1 MAG2 MAG MAG MAG MN device HMAG Fig. 5 An example of CSPMIPv6 inter-structure for load balancing movement 4.1 Load balancing mechanism for clustered PMIPv6 domain (LB-CPMIPv6) In this proposed mechanism, a load balancing mechanism is developed for a cluster PMIPv6 to improve the efficiency of MNs and accordingly the overall system Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 11 of 23 MAGi HMAGk LMA Start Start Start Set initial values for Set initial values for Set initial values for s f, µ f , s,Nm , L, fL No No No On Packets On Packets On Packets Receiving receiving Receiving Yes Yes Yes No No No pi > i pt > pT > Yes Yes Yes No No f< f fL< fL MAGi preforms a load balancing action Yes Yes MAG select the handover MNs (HMN) HMAG sends load balancing request message LMA sends load balancing request message to MAGi to HMAG MAG selects the best MAG from the non- overloaded candidate MAGs MAG obtains the MAGs load from the Heartbeat message exchanges among the MAGs MAGi sends handover command to HMNs to change their routing to the TMAG MAG sends load status to HMAG periodically i K HMAG sends load status to LMA periodically Fig. 6 Load balancing operations within CSPMIPv6 domain performance is improved. This is due to the need to and will hereafter be denoted by λ where the N is the i m take into consideration the intra- and inter-domain mobil- number of the measurements. The N measurements are ity during the load balancing process. The MAG located used to estimate the λ for MAG , which is computed as within the CSPMIPv6 domain acts as the gateway between theaverage arrivalrateatacertaintimeinterval. After the MNs and the HMAG. Thus, the MNs must be con- that, the MAG calculates the average packet arrival rate nected to the MAG to be connected to the network. using the weighted moving average technique under the Subsequently, the MAG could become overloaded if the assumption that μ istheaverage servicerateoftheMAG , i i number of the connected MNs increases in the net- whichisusedby[13] and is mathematicaly expressed as work. This constraint has motivated, a new load balancing follows: mechanism, which is applied to reduce the load at mainly the overloaded MAGs. (N − j + 1)λ m N −j+1 j=1 λ = (1) In order to ensure the standardization of the per- j=1 formance analysis as a comparative platform, the LB- CPMIPv6 mechanism performance analysis has reused Thereasontoutilizetheweightedmovingaverage the parameter values and assumptions that have been pre- method is to reveal the uncontrolled action. In addition, sented byKimand Leein[13]. Hereafter, the proposed it gives a higher weight to the current traffic sample as load balancing mechanism for the PMIPv6 network in [13] compared to the old traffic sample in the measurement as should be referred as the “LBM-PMIPv6 mechanism” for proposed in [27]inorderto computetheMAGloadpre- the sake of simplicity. In the CSPMIPv6 system model, the load at MAG is cisely. Then, the p can be expressed as where the λ is i i measured according to the average packet arrival rate in a theaverage arrivalrateand the μ is the average services particular interval time. The similar measurement is used rate at a certain time. By considering the MAG process- th for measuring the arriving rate at a certain j time interval ing capacity into account, θ is the maximum acceptable i Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 12 of 23 load on the MAG i.If p >θ ,MAG becomes overloaded the HMAG will send a heartbeat message to its related i i i k and will perform a load balancing technique. MAGs to inform the most overloaded MAGs to perform In the LB-CPMIPv6 mechanism, every MAG is sup- a load balancing action. In this work, the extended field (F posed to send its load and its domain number in a flag and load status field ) in the heartbeat message that is periodical manner to the related HMAG using the heart- given by [13]isreusedinthe sameway. beat message, which has been introduced in [28]. The Finally, every HMAG in the CSPMIPv6 domain has to heartbeat message is exchanged periodically by the MAG send its load that is received from the related MAGs to the information related to their related HMAGs within the LMA. This is performed using the heartbeat message to CSPMIPv6 domain. This is done to inform the HMAGs compute the overall system performance and operates as with their loads as well as to detect the reachability of follows. the other end links. In this work, the heartbeat message Once the HMAG receives the total loads p from each is extended to include the load status and the domain related MAGs, the HMAG send these loads to the related number, as shown in Fig. 7. In addition, the Proxy Bind- LMA periodically using the heartbeat message. ing Acknowledgment (PBA) and the Local Proxy Binding Upon receiving the p loads from the associated Acknowledgment (LPBA) messages, which are presented HMAGs, the LMA computes the load status F1, using in [2]and[4] respectively are extended, as shown in the received loads to measure the overall system perfor- Figs. 8 and 9. This extension is to include the domain mance within the CSPMIPv6 domain. The total load can number during the initial attachment of the MAGs while be expressed as: in the other control signaling the domain number is set to be zero. These two mechanisms are used to define p = p T t the domain numbers, which are dynamically done by the k=1 LMA according to the number of its related domains or statically done during the installation. where the N isthenumberoftheHMAG in theCSP- MIPv6 domain and p is the total load at the HMAG . t k i=1 After that, the LMA in the CSPMIPv6 domain measures f = (2) 2 the F1 according to the MAGs load received by the related M ∗ ( p ) i=1 HMAG in the entire system as expressed in Eq. (2), where Each domain number and loading information that is M denotes the number of HMAGs in the system. received by the HMAGs, LMA, and MAG are saved If f is less than θf , a heartbeat message request is sent in their databases, which are then used in the intra- by the LMA to the most overloaded HMAGs to inform and inter-domain load balancing. Subsequently, every them to perform a load balancing action. The LMA uses HMAG also measures its load status, F1, via employing k the received loads from all the related HMAGs to deter- the stored load information on its database that is received mine the most overloaded HMAGs. A new flag is also from its related MAGs within its domain. The total load added to the heartbeat message is named F andisset to1if for each HMAG domain p can be expressed as p = t t the heartbeat message comes from the LMA entity to the p where the M is the number of the MAGs within related HMAGsorzeroifthe heartbeatmessage comes i=1 the HMAG domain in the CSPMIPv6 domain. If p >θ, k t fromtheHMAGtotherelated MAGs. the HMAG measures the Fairness Index (F ) according to k I Once the overloaded HMAG receives the heartbeat the Eq. 2 that is given by [29]. The F lies between 0 and 1. message, a load balancing action is performed by sending If all the MAGs within the HMAG domain have the same a heartbeat message to the related MAGs, which in turn load, the F is 1. selects the HMNs in the overlapped area. The HMNs must Subsequently, the HMAG uses the MAGs load infor- change their point of attachment to another MAG. The mation, which is stored in its policy database to com- criteria of the HMN selection are discussed in the next pute the f value and the compares it with θf.If f <θf, subsection. 01 2 3 0 1 2 3 45 6789 0123456 7 8 9 012345 67 8 9 0 1 Reserved L F U R Sequence Number Heartbeat Time Interval Load Information Domain Number Fig. 7 Heartbeat message, including the domain number Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 13 of 23 01 2 4 0 1 2 3 4 5678901 23 4 5 6 7 8 9 01 2 3 456789 01 Sequence Number AHLKMRPSNQReserved Life Time Domain Number Mobility Options Fig. 8 PBA message, including the domain number According to the above explanation, the LB-CPMIPv6 After the HMNs selection by the MAG from its candi- mechanism only affects the protocols with multiple date list, the MAG now is ready to choose the best target domains. In other protocols, there is no impact because MAG to where the HMNs bind. The selection of the tar- the LB-CPMIPv6 mechanism deals with the whole get MAG is performed as described in the subsection of domain as one domain. Thus, selecting the MAG will be the target MAG selection. performed based on the LBM-PMIPv6 mechanism [13]. 4.1.2 Target MAG selection 4.1.1 HMN selection For an enhanced load distribution, the selection of the tar- The wrong selection procedure of the HMNs candidate get MAG must be performed as accurately as possible. significantly degrades the performance of the system. Therefore, the Technique for Order Preference by Sim- Thus, the system must be adapted to appropriately choose ilarity to Ideal Solution (TOPSIS) algorithm is modified the HMNs in a better manner. In this work, the process to determine the target MAG in this research. Further- of HMNs selection by the overloaded MAG is performed more, in this study, additional operations are proposed to based on some criteria as follows. adopt the algorithm with the clustered PMIPv6. The intra- The MAG chooses the HMN that has an option to and inter-cluster handoffs within the CSPMIPv6 domain change its connection. This indicates that the HMN is have been considered due to the adaptation process. The located between different networks that are advertising system has achieved better selection technique to the tar- their services to such HMN in order to maintain a contin- get MAG among the candidates MAGs, which have been uous IPv6 session. This is achieved if the HMN is located reflected in the system performance in terms of handover in the MAGs overlapped area and it receives a Signal latency, end-to-end delay and queuing delay as presented Strength (SS) from all of them. in Section 6. The enhanced processes of selecting the best The MAG creates a candidate list for the HMNs that target MAG are performed as follows: exist in its overlapped and receive an RSS from another The MAG utilizes the RSS, which is reported by the MAG. HMN to determine the available network for the HMN. Then, the MAG should select the HMNs that require The MAG then places the available networks as candi- the highest data rate from the candidate HMNs. date networks (i.e., candidate MAGs). This is performed TheMAGshouldnotselectthe HMNsthat havea to select the best candidate MAG in terms of RSS, load real-time connection (e.g., audio and video) due to the status and the domain number during the load balanc- sensitivity to delay leading to increase the handoff latency. ing action. The technique for order preference is adapted To determine the non-real HMN session by the MAG, from the TOPSIS algorithm that is presented in [27]. The the “Traffic Class” or the “Flow Label” field of IPv6 must TOPSIS algorithm is used by the serving MAG to deter- be examined by the MAG for all the candidate HMNs mine the target MAG. This algorithm used to choose the in the MAG overlapped area [30]. So disturbing the real- optimal MAG as possible according to the Signal Strength time session during the handover latency will be avoided. reported by the HMNs and the load status of that MAG. If the MAG overlapped area does not contain any non-real Observation shows that the TOPSIS algorithm is not suit- HMN session, the HMNs with the highest data rate will be able when applied within the clustered PMIPv6 domain. selected. This is due to the fact of dividing the domains into local 01 2 3 0 1 2 345 6789 0 1 2345 67 8 9 0 1 2345 67890 1 Status K R P S N Q Reserved Sequence # Life Time Domain Number Mobility Options Fig. 9 LPBA message, including the domain number Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 14 of 23 sub-domains, which is not considered in the TOPSIS with the normalized signal strength. For the sake of sim- algorithm, subsequently increasing the communication plicity the wp ¯ is replaced by v ,wherethe wS is replaced i p i overhead. Therefore, some enhancement has been imple- by v mented by the LB-CPMIPv6 mechanism starting with Step 5: Determine the optimal network C (min v ,max modification of the exchanges of messages between the v and the worst network C (max v ,min v )from V S p S system entities until of change the HMN its points of matrix: attachment to ideal TMAG. The TOPSIS algorithm steps ∗ ∗ ∗ C = v , v C ={v ˆ , v ˆ }, ∀ ∈{1, 2, ... , n}.(9) including the additional operations of selecting the opti- p S i p S mal target MAG for the HMNs are described as follows: Step 6: The TOPSIS algorithm calculates the separation Step 1: The TOPSIS algorithm constructs the decision measures by using the Euclidean distance. The separation matrix D, which is a [1 × n] matrix, as: of each candidate MAG from the optimal and the worst D = {C , C , ... , C },(3) 1 2 n MAG, S and C are calculated as: th where C represents a pair of p and S for the i candidate i i i ∗ ∗ ∗ 2 S = v − v (v − v ) , ∀ ∈{1, 2, ... , n}. (10) p S i i p MAG in the matrix D,as: C = (p , S ) , ∀ ∈{1, 2, ... , n},(4) i i i i ˆ 2 S = v − v ˆ (v − v ˆ ) , ∀ ∈{1, 2, ... , n}. (11) i p p S S i where p denotes the load status and S represents the sig- i i nal strength of candidate MAG and n isthenumberof Step 7: Next, the MAG ranks the preference order after candidate MAGs in the matrix D. Moreover, the TOPSIS calculating the relative separation measure as: algorithm is designed to avoid the candidate MAGs, which ˆ ˆ load is more than a predefined threshold θ. x = S /S + S , ∀ ∈{1, 2, ... , n}. (12) i i i i Step 2: The TOPSIS algorithm computes the normalized Step 8: This step is performed according to two different decision matrix D, which is a [ 1 × n]matrix, as: cases. First, if the closest MAG to the ideal network envi- ¯ ¯ ¯ ¯ D = C , C , ... , C,(5) 1 2 n ronment has the same domain of the serving MAG, the serving MAG selects it without any hesitation as the tar- ¯ ¯ where C represents a pair of p ¯ and S for the each i i i get MAG. Second, if the closest MAG to the ideal network candidate MAG in the matrix D,as: is from a different domain, the serving network looks into ¯ ¯ C = p ¯ , S , ∀ ∈{1, 2, ... , n},(6) the ideal network candidate MAGs list to see if there is i i i i another ideal one has the same domain in order to choose it instead of the ideal one has a different domain. The where p ¯ = p ¯ / p ¯ and denotes the value of normalized i i i=1 selection of the TMAG from the ideal network candidates ¯ ¯ ¯ maintains the system stability in terms of signal strength load status, while the S = S / S and represents the i i i=1 and the load status. This means, the process of selecting value of normalized signal strength of candidate MAG the TMAG that have the same domain will not affect the and n is the number of candidate MAGs in the matrix D. thresholds of the signal strength and the load, which will Step 3: The TOPSIS algorithm uses the weight value (w), lead to the maintenance of the connection session without which is a system parameter. This weight (w) and its com- any service disruption. plement to 1 (1 − w) are used to weight the p and S , i i Step 9: When the serving MAG has done the selection of respectively. Since the load at the candidate MAG (p ) has the closest MAG to the ideal network environment from higher priority than the signal strength (S ), the weight the ranking preference, the serving MAG starts a load value (w) have to always be greater than 0.5. balancing process by sending a handover command mes- Step 4: The TOPSIS algorithm calculates the weighting sage to HMN. In order to select the ideal TMAG, a new decision matrix: V,as: load balancing signaling is proposed. In the next section, the new load balancing signaling within the clustered ¯ ¯ ¯ V = wC , wC , ... , wC,(7) 1 2 n protocols is explained in detail. ¯ ¯ where wC represents a pair of wp ¯ and wS for the each i i i candidate MAG in the matrix D,as: 5 Load balancing signaling for the clustered PMIPv6 domain ¯ ¯ wC = wp ¯ , wS , ∀ ∈{1, 2, ... , n},(8) i i i i As mentioned earlier, every MN within the PMIPv6 where the wp ¯ is the result of the multiplication of the domain must connect to a particular MAG to commu- weight value with the normalized load status value and the nicate with other MNs. This can lead to overloading the wS is the result of the multiplication of the weight value MAG, especially in large networks, when the number of i Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 15 of 23 MNs is substantial. This section presents the required sig- When the MAG receives the handover report from naling of our proposed work to mitigate the load of the the HMN, the MAG extracts the IDs of the candidate overloaded MAG. Figure 10 shows that this signaling is MAGs, which is given by the report and subsequently extended from the CSPMIPv6 signaling framework, which stored as a list on its policy database. Then, the MAG includes the inter- and intra-handover signaling opera- utilizes the Handover initiates (HI) and the Handover tion [4]. Given that, the MN sends a report to the serving Acknowledgment (HAck) messages, which are introduced network, which includes the MN-ID, MN-IID, the new by [11] and extended by [13] to determine the load of MAG-ID,andtheRSS.Thisreportissentonlyifthe RSS the candidate MAGs. This is done by sending an HI exceeds a threshold as performed in Third Generation message by the MAG to all candidates MAGs includ- Partnership Project (3GPP). The scenario of performing a ing the MN-ID. Once the HI message received by the load balancing handover in the clustered PMIPv6 domain candidate MAGs, the candidate MAGs reply by send- is performed as follows. ing HAck messages emulated to the MAG including The MAG performs a load balancing procedure accord- the current loads information. A flag N used as intro- ing to three cases. Therefore, when a load of MAG duced by [13] to distinguish between the HI and HAck exceeds a specific threshold or if its cluster head HMAG messages. sends a load balancing request or if the LMA orders The MAG employs the received loads information from the HMAG , to perform a load balancing, which in turn thecandidateMAGsandthedomainnumbertochoose moves this order to the related MAG .Afterthat,the the TMAG, which the HMN will move to as described MAG starts to perform a load balancing procedure by in Section 2.2. Subsequently, a Handover Command Mes- choosing theappropriateHMNfromtheoverlappedarea sage (HCM) is sent by the MAG to the HMN. to perform a handover action. This HMN selection is Upon receiving the HCM, the HMN starts a new link performed as presented earlier. connection with the TMAG. HMAG2 HMN SMAG HMAG1 C-MAG1 C-MAG2 LMA Report (MN-ID, MN-IID,RSS, New MAG Info.) Tunneling HI Message HI Message Determine the MAGs HAck Message (C-MAG1 Load Info. Load HAck Message (C-MAG2 Load Info. Determine the TMAGs for the HMN Handover Command(MAG1-Info.) Establish L2 connection by the MN MN Starts New L2 Connection LPBU Message LPBA Message Tunneling Router Solicitation (RS) Router Advertisement(RA) DATA Handover Command(MAG2)) Establish L2 Connection by The HMN MN Starts New L2 Connection LPBU Message PBU Message Updating BUL PBA Message LPBA Message Updating BUL Router Solicitation (RS) Tunneling Router Advertisement (RA) DATA Fig. 10 Load balancing signaling in the CSPMIPv6 domain Inter-cluster Loab balancing Intra-cluster Load balancing Extra Signaling for LB-CSPMIPv6 Updating BCE Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 16 of 23 Table 1 Parameters for experimental results When the TMAG detects the attachment of the HMN, Parameter Values an LPBU message sends to the HMAG . Once the HMAG receives the LPBU message, the k θ 0.925 HMAG fetches its binding table by searching for the θ 0.105 HMN information. If the HMN moves to another MAG θ 0.5 within the HMAG cluster domain, which is as called θ 0.5 intra-cluster handoff mobility, the HMAG updates its θ 0.175 fL Binding Cash Entry (BCE) by setting the New MAG (NMAG) address instead of the TMAG address in its μ 2 packet/ms MAG field. On the other hand, if the HMN moves to K 20 another MAG in a new HMAG (NHMAG) cluster domain Packet size 1024 KB within the CSPMIPv6 domain, which is as called inter- cluster handoff mobility, the NHMAG will not find any matching entry in its BUL table. Since the HMN comes from another cluster, so the NHMAG sends PBU to the the network simulator NS2 [32, 33]. As illustrated in LMAtoupdateits BCEfor theHMN.Then, theHMAG Fig. 11, which represents the system topology, the MNs field entry in the BCE of the LMA will be modified by communicate with each other including the CN that is setting the NHMAG address and releases the old one for connected to the CSPMIPv6 domain through the core net- the HMN. After that, the LMA replies to the NHMAG via work. The Poisson process has been used by CN and MNs sending a PBA message. to generate packets with a mean rate of 2 packets/ms. The NHMAG, upon receiving the PBA message sends Each MAG measures the packet arrival rate (λ)atevery an LPBA message to the NMAG after updating its binding 50 ms. When the number of measurements reaches 20 entry for the HMN. times by the MAG, the MAG calculates the λ using the The NMAG consequently updates its BUL for the HMN weighted average method as depicted in Eq. (1). In the and advertises the HNP to the HMN. proposed network topology, the MNs are randomly scat- tered over 10 MAGs. In addition, the MAGs are equally connected with two HMAGs (HMAG1, HMAG2) while 6 Performance evaluation the HMAGs are associated to one LMA. The load status This article presents the development of a load balancing foreachMAGiscarried by theheartbeat message thatis mechanism among MAGs in the CSPMIPv6 domain. The sent every single second to their related HMAG. The time performance analysis of this algorithm and the compara- between the first Heartbeat and the next should be small tive analysis has been done using discrete event simulation as recommended in [28]. Similarly, the HMAGs send their and in particular the NS2. For the evaluation purposes, load status to the LMA to measure the overall system per- the work in [13] is re-implemented and the CSPMIPv6 formance. Several MNs are scattered randomly over the protocol, which is presented in [4]. Also, this work reuses MAGs areas. The total load varies between 0.05 to 0.8. the parameter values and assumptions that have been Every MAG is associated with a limited queue K as men- used in [13] to ensure a level of comparative platform, tioned in Table 1.Forsimplicity,eachMAG inthenetwork as shown in Table 1.Table 1 shows the new setup val- topology is considered to have same service rate μ and ues needed by LB-CPMIPv6 mechanism for performing a threshold value θ . The simulation is conducted under load balancing. In addition, the performance gain is calcu- three different scenarios. In the first scenario, the LB- lated as in [31] to show the variation results between the CPMIPv6 mechanism is applied in the PMIPv6 network proposed LB-CPMIPv6 and LBM-PMIPv6 mechanisms (without-clustering) and is compared with LBM-PMIPv6 based on Eq. (13), where x and x ´ represent the results pro- mechanism [13]. This scenario can show the impact of LB- duced using LB-CPMIPv6 and LBM-PMIPv6 mechanisms CPMIPv6 mechanism in the PMIPv6 protocol that uses respectively. no clustering technique. In the second scenario, each MN can connect to one, two or three MAGs with an equal (x − x ´) probability within the CSPMIPv6 domain. This scenario Performance gain =  × 100 (13) (x ´) is performed to inject the overlapped area with a high number of MNs in order to show the actual impact of the intra-cluster handoff process on the system performance. 6.1 System setup In the last scenario, each MN is connected to one or zero This section illustrates the simulation setup used in the MAG with a probability equal and 0.2 and 0.8 respec- experiments in order to evaluate the proposed load bal- tively, without any concentration on the overlapped area ancing mechanism. The experiments are performed using between the MAGs. Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 17 of 23 LMA CN MAG MAG MAG MAG HMAG2 HMAG1 MAG MAG MAG MAG MAG MAG Fig. 11 The network topology used for simulation 6.2 Results and discussion from HMN. In this case, the LB-CPMIPv6 mechanism In this section, the performance metrics evaluated are works precisely similar to LBM-PMIPv6 mechanism, and the average queuing delay, handover latency, transmission PMIPv6 protocol shows the highest average queuing delay since it does not support the load balancing. Thus, the rate and the packet loss. The queuing delay is defined as the summation of the waiting time of each packet in packet service time inside the overloaded MAG buffer the queue per MAGs. The transmission rate is measured increases as a result of traffic from connected MNs. by calculating the average amount of data transmission In the second scenario, Fig. 13 illustrates the average from the MAGs for the entire simulation. Three scenarios queuing delay per packet at the MAG versus the overall are conducted to demonstrate the enhancement of LB- system load for the entire simulation time within a clus- CPMIPv6 mechanism in the clustered protocol against tered environment. In this scenario, the CSPMIPv6 proto- the LBM-PMIPv6 mechanism in addition to the scenario col is employed as an experimental environment to show situation of no load balancing. The scenarios have been the effectiveness of the proposed LB-CPMIPv6 mecha- conducted as follows. nism in the clustered domain, which totally differs from In the first scenario, the proposed mechanism applied the first scenario applied in the non-clustered domain. on the standard PMIPv6 protocol, which means that It is evident that the LB-CPMIPv6 and LBM-PMIPv6 the MAGs belong to the same domain (no hierarchical mechanisms outperform the average queuing delay of the domain). In this scenario, a comparison between LB- CSPMIPv6 protocol, which has no load balancing. Per- CPMIPv6 mechanism and LBM-PMIPv6 mechanism in forming the load balancing action leads to the reduction PMIPv6 domain is carried out in terms of measuring of the packets buffering time at the overloaded MAGs. average queuing delay. The result of this comparison is Furthermore, the performance for LB-CSPMIPv6, LBM- depicted in Fig. 12. As observed from the figure both, PMIPv6 and CSPMIPv6 shows same results regarding the the LB-CPMIPv6 mechanism and LBM-PMIPv6 mech- queuing delay before the predetermined thresholds take anism have achieved similar results in terms of average place. queuing delay, while PMIPv6 protocol, which has no load When the system load reaches 0.105 and 0.175, the balancing mechanism, shows a higher average queuing HMAG and/or the LMA starts to perform a load balanc- delay. In the LB-CPMIPv6 mechanism, if there is only ing action because the values in the figure are influenced one domain (no clustering), the serving MAG selects the after the predefined thresholds take place. In addition, TMAG according to its load and based on RSS reported when the total load reaches 0.35 for θ = 0.5, a few f Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 18 of 23 LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 12 The average queuing delay per MAG in the PMIPv6 protocol MAGs suffer from the overload issue. Subsequently, the affects the intra-cluster handoff. The performance gain of overloaded MAGs start performing a load balancing pro- LB-CPMIPv6 mechanism over LBM-PMIPv6 mechanism cedure as depicted in Fig. 13. The LB-CPMIPv6 mecha- is 15.66% on average. nism shows identical results regarding the queuing delay The LB-CPMIPv6 mechanism selects the ideal MAG in comparison to the LBM-PMIPv6 mechanism before from the best candidates through employing the domain reaching the respective thresholds. On the other hand, number, which in turn leads to shortening the time when the respective thresholds are reached, the average needed for delivering packets to their destination after queuing delay increases whenever the load increases. This the handoff. For example, when the load reaches 0.105 isduetotheincrease of thenumberofexchanged pack- on the HMAG, the HMAG performs load balancing ets among the MNs. However, the LBM-PMIPv6 mecha- by sending a message to the overloaded MAG, which nism shows high queuing delay compared to LB-CPMIPv6 is located on its domain and each overloaded MAG mechanism, as illustrated in Fig. 13.Thisisbecause sub- selects the TMAG based on its load, domain number and domains are not utilized in the clustered protocols, which the RSS. LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 13 The average queuing delay obtained from scenario 2 Avg. queuing delay (ms) Avg. queuing delay (ms) Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 19 of 23 Figure 14 illustrates the packet loss ratio between LB- different domain, which increases the buffering time. This CPMIPv6, LBM-PMIPv6 mechanisms and the CSPMIPv6 results in the increasing of congestion from the queuing protocol. The advantages of applying the load balancing system perspective. mechanisms in the protocols can be demonstrated clearly A related example to this is the movement of the HMN to an overall reduction in the overloaded MAGs. For to another cluster domain causes an extra delay due to example, the load balancing reduces the increased level the time needed to exchange its information among the of buffer utilization, which in turn reduces the number HMAGs. After this, this information should be emulated of lost packets. In addition, the LB-CPMIPv6 mechanism to the TMAG, which in turn needs a buffering technique achieves better results in terms of packet loss as com- to preserve the packets during the handover process. pared to the LBM-PMIPv6 mechanism. This is because Figure 15 compares the effects of LB-CPMIPv6 mech- the fact that the LB-CPMIPv6 mechanism moves the anism and LBM-PMIPv6 mechanism on the handover HMNs within the same domain as possible, which results latency. The handover latency is the interval between the to bring down the time of the handoff process leading to time of the last packet that is received by the HMN from the reduction packet loss. In other words, shortening the the old path and the time of the first packet that is received time needed to perform the handoff process leads to the from the new path by the HMN. The LB-CPMIPv6 reduction of the packet waiting in the buffer, which in turn mechanism outperforms the LBM-PMIPv6 mechanism in reduces the packet loss. terms of the handover latency. This is due to that the When the total load reaches 0.175, the LMA and a min- TMAG is selected based on the domain factor. In other imum one of the HMAGs exceed their thresholds, which words, the time needed to perform a handoff process are depicted in Table 1 (θ and θ), and for that the bal- by the HMNs is reduced. This is done by eliminating ancing function will be triggered by LMA and/or the the authentication process on TMAG and performing the respective HMAG. The TMAG that has taken the same handover without the involvement of the LMA, which domain with the MAG that has load balancing action will may be located far from the HMNs. The handover pro- be selected according to the MAG selection criteria in cess is performed by the overloaded MAG if the load in LB-CPMIPv6 mechanism. This definitely leads to short- the LMA, HMAG or MAG exceeds their predetermined ening the time needed to register the HMN on the TMAG, thresholds. As in Fig. 15, the handover is started, when the which in turn decreases the packet waiting time in the total load reaches 17.5 and is performed again when some MAGs become overloaded. The performance gain of LB- queue. This enhancement leads to the reduction of the packet congestion from the point of view of the queu- CPMIPv6 mechanism over LBM-PMIPv6 mechanism is ing system. Moreover, selecting the TMAG based on the almost 32.68%. RSS and load status only in the LBM-PMIPv6 mechanism In the third scenario, the average queuing delay, trans- increases the packet loss ratio. This is due to the long reg- mission rate and end-to-end delay are measured, as shown istration time that is needed to register the HMN in a in Figs. 16, 17,and 18 respectively. LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 14 Number of packet loss for LB-CPMIPv6 mechanism, Kim and Lee [13] and non-load balancing Packet Loss Ratio (%) Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 20 of 23 LB-CPMIPv6 mechanism LBM-PMIPv6 [13] 4.5 3.5 2.5 1.5 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 15 Impact of intra-cluster handoff on the handover latency Figure 16 depicts the impact of the LB-CPMIPv6 on The performance gain of LB-CPMIPv6 mechanism over the average queuing delay in comparison with the LBM- LBM-PMIPv6 mechanism is almost 9%. PMIPv6 mechanism and the original CSPMIPv6 proto- Figure 17 shows the average data transmission rate from col. The figure shows that the LB-CPMIPv6 mechanism the MAGs per MNs in the third scenario. The MNs increases the performance of the LBM-PMIPv6 mecha- scattered randomly within the CSPMIPv6 domain. It is nism even when the overlapped area is characterized by a obvious that the LB-CPMIPv6 mechanism has a higher small number of MNs. This can be attributed to the pro- data transmission rate than the other mechanisms, while posed mechanism performance to select the TMAG from the CSPMIPv6 with no-load balancing has the least data the same domain when the overloaded MAG performs a transmission rate. load balancing action. This definitely leads to the short- We observed from Fig. 17 that the data transmission rate ening of the time needed to register and authenticate the is roughly stable in the LB-CPMIPv6 and LBM-PMIPv6 MN on the TMAG, which in turn decreases the packet mechanisms. However, in the case of no-load balancing, waiting time in the queue, especially in the limit queue. the data transmission rate decreases whenever the MNs LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 16 The average queuing delay obtained from the third scenario Avg. queuing delay (ms) Avg. Handover Latency (ms) Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 21 of 23 0.9 0.8 0.7 0.6 0.5 0.4 LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0.3 0 100 200 300 400 500 600 MN’s arrival rate to the MAGs Fig. 17 The transmission rate obtained from the third scenario arriving rate increases. This is due to the absence of load the HMN and authenticating it in the TMAG, which in balancing that leads to an unbalanced situation at the turn increases the amount of sent packets to their targets. MAGs within the CSPMIPv6 domain. Furthermore, LB- However, in the LBM-PMIPv6 mechanism, the traffic usu- CPMIPv6 mechanism shows a significant enhancement in ally has an extra delay as a result of sending packets to data transmission rate compared to LBM-PMIPv6 mech- another cluster. anism, as shown in Fig. 17. This is due to the fact that Figure 18 presents the measured average of the end- the LB-CPMIPv6 mechanism gives higher priority for to-end delay per MAG in the CSPMIPv6 versus the total the selection of the TMAG based on its domain without load on the overall system. Interestingly, the LB-CPMIPv6 affecting the load status or the SS threshold. This mech- mechanism outperforms the LBM-PMIPv6 mechanism anism increases the traffic among the MAGs resulting in and the CSPMIPv6, which has no load balancing, despite increasing the data transmission rate. Furthermore, for- reducing the overlapped area. This is due to performing warding the HMNs traffic to the TMAG that is located in load balancing action by the overloaded HMAGs, which the same cluster reduces the time needed for registering leads to distribute the load within their clusters. This, LB-CPMIPv6 mechanism LBM-PMIPv6 [13] no-load balancing 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Load Fig. 18 End-to-end delay per MAG versus the total load Avg. End-To-End delay (ms) Avg. Dtata Rate/MN from MAGs (Mbps) Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 22 of 23 in turn, leads to shortening the routing path as well as Acknowledgments Thanks to all my family, Sons, friends, and my colleagues who helped me to choosing the most closest MAG to the serving MAG. achieve this work. A special thanks to my supervisor Dato Prof. Shamala Selecting the TMAG based on its domain number has Subermaniam. a positive impact on the overall system performance. Availability of data and materials Subsequently, the result shows that the utilizing domain NS2 simulator and PMIPv6 patch are employed to test the work in this paper. number during the TMAG selection reduces the end- to-end delay due to the path faster recovery for the Authors’ contributions SMG designed the methods, conducted the experiments, evaluated HMN after the handoff. Furthermore, the original CSP- performance, and wrote the paper. SS defined the research area, problems MIPv6 protocol has a higher end-to-end delay due to and objectives. ZAZ and AM have equally contributed by given the final the unfair distribution of load. The unbalancing MAGs approval of the version to be published and contributed to the analysis. All authors read and approved the final manuscript. suffer from the heavy load, which in turn increases the overhead on the MAGs queue, causes extra packet time Authors’ information delay. Moreover, CSPMIPv6 tends to have relatively Safwan Ghaleb has received his bachelor degree in Computer Science from University of Jordan, Amman, Jordan, in 2009, the master degree in computer long paths, which also contributes to increasing the science from Jordan University of Science and Technology, Irbid, Jordan in end-to-end delay. 2012. He is working towards Ph.D. in computer networks, Universiti Putra Malaysia. His research interest include Internet of Things (IoT), Wireless and Mobile Networks, and Data Mining. 7Conclusions S. Shamala received the B.S. degree in Computer Science from University Putra PMIPv6 protocol and its extensions have been proposed Malaysia (UPM), in 1996, M.S. (UPM), in 1999, Ph.D. (UPM) in 2002. Her research to provide a seamless handover action within a localized interests are computer networks, simulation and modeling, scheduling and real time system. Dr. Shamala is now Prof. at the Department of management network. This is achieved via relieving the Communication Technology and Networks, Faculty of Computer Science and MN from any signaling-related to the mobility process Information Technology, University Putra Malaysia (UPM), Malaysia. when the MN changes its link. This is done by adding the Dr. Zuriati Ahmad Zukarnain is a professor at the Faculty of Computer Science and information Technology, University Putra Malaysia. She has served as a new MAG that performs the mobility related-signaling Head of Department of Communication Technology and Networks at the with the LMA instead of the MN. Furthermore, the MAG Faculty of Computer Science and information Technology, University Putra establishes a tunnel with LMA to send and receive the Malaysia. She received her PhD from the University of Bradford, UK. Her research interests include: Efficient multiparty QKD protocol for classical packets of the MN. However, to establish a new link con- network and cloud, load balancing in the wireless ad hoc network, quantum nection, the MN has to be associated with a specific processor unit for quantum computer, Authentication Time of IEEE 802.15.4 MAG. This association could overload the MAG. Conse- with Multiple-key Protocol, Intra-domain Mobility Handling Scheme for Wireless Networks, Efficiency and Fairness for new AIMD Algorithms and A quently, the LB-CPMIPv6 mechanism has been proposed Kernel model to improve the computation speedup and workload in this article to fairly distribute the loads among the performance. She has been actively involved as a member of the editorial MAGs fairly. The main advantage of LB-CPMIPv6 is its board for some international peer-reviewed and cited journals. Dr. Zuriati is currently undertaking some national funded projects on QKD protocol for capacity to consider clustered domain within the clustered cloud environment as well as routing and load balancing in the wireless ad protocols, which is not considered in other competitive hoc network. Dr. Zuriati is the founder of ZA Quantum Sdn Bhd, the start up mechanisms. company from University Putra Malaysia to produce a software designing tool for Quantum Communication known as Quantum Communication Simulator In the LB-CPMIPv6, the HMN that has a real-time ses- (QuCS). sion will not be selected during the process of the load Abdullah Muhammed received the bachelor degree in computer science from balancing; this restriction relieves the critical applications Universiti Putra Malaysia in 1998, the master degree in computer science from Universiti Malaya in 2004 and the PhD degree in computer science from the from service disruption. Furthermore, the CSPMIPv6 University of Nottingham, United Kingdom in 2014. He is a senior lecturer at handover signaling has been extended to be adapted with the Department of Communication Technology and Networks, Faculty of the newly proposed load balancing mechanism. Moreover, Computer Science and Information Technology, Universiti Putra Malaysia and is currently the HoD. His main research interests include cloud computing, the LPBA, PBA, and the heartbeat messages are modified mobile and wireless network, scheduling, heuristic and optimization. to enable sharing of the domain number for the new load balancing mechanism. Competing interests The authors declare that they have no competing interests. The LB-CPMIPv6 mechanism is implemented and simulated using the well-known NS2 simulator. The Publisher’s Note evaluation of the LB-CPMIPv6 mechanism in com- Springer Nature remains neutral with regard to jurisdictional claims in parison to the LBM-PMIPv6 load mechanism and published maps and institutional affiliations. CSPMIPv6 protocol is performed in terms of queu- Author details ing delay, packet loss ratio, end-to-end delay, and Department of Communication Technology and Network, Universiti Putra transmission rate. The results show that the new 2 Malaysia, 43400 Serdang, Selengor D.E., Malaysia. Sports Academy, Universiti load balancing mechanism achieves a better perfor- Putra Malaysia., 43400 Serdang, Selengor D.E., Malaysia. mance by reducing the average queuing delay, packet loss, end-to-end delay, and increasing the transmission Received: 20 October 2017 Accepted: 27 April 2018 rate. Ghaleb et al. EURASIP Journal on Wireless Communications and Networking (2018) 2018:135 Page 23 of 23 References Proxy Mobile IPv6 networks: an IP multicast perspective (IEEE, Honolulu, 1. PCD Johnson, J Arkko, Mobility support in IPv6. Technical report, RFC 2014), pp. 766–770. https://doi.org/10.1109/ICCNC.2014.6785433 3775 (June 2004). https://www.rfc-editor.org/info/rfc3775 23. S Jeon, RL Aguiar, N Kang, Load-balancing Proxy Mobile IPv6 Networks 2. ELKDVCKS Gundavelli, B Patil, Proxy Mobile IPv6. Technical report, IETF, with mobility session redirection. IEEE Commun. Lett. 17(4), 808–811 RFC 5213 (2008). https://www.rfc-editor.org/info/rfc5213 (2013) 3. SM Ghaleb, S Subramaniam, ZA Zukarnain, A Muhammed, Mobility 24. C-M Huang, M-S Chiang, PB Chau, in 2015 IEEE International Black Sea management for IoT: a survey. EURASIP J. Wirel. Commun. Netw. 2016(1), Conference on Communications and Networking (BlackSeaCom).A 165 (2016) load-considered fast media independent handover control scheme for 4. AJ Jabir, SK Subramaniam, ZZ Ahmad, NAWA Hamid, A cluster-based Proxy Mobile IPv6 (LC-FMIH-PMIPv6) in the multiple-destination proxy mobile IPv6 for IP-WSNs. EURASIP J. Wirel. Commun. Netw. 2012(1), environment (IEEE, Constanta, 2015), pp. 171–175. https://doi.org/10. 1–17 (2012) 1109/BlackSeaCom.2015.7185109 5. JH Lee, KD Singh, JM Bonnin, S Pack, Mobile data offloading: a host-based 25. SM Raza, D Park, Y Park, K Lee, H Choo, in Proceedings of the 10th distributed mobility management approach. IEEE Internet Comput. 18(1), International Conference on Ubiquitous Information Management and 20–29 (2014) Communication. IMCOM ’16. Dynamic load balancing of local mobility 6. MM Islam, E-N Huh, Sensor proxy mobile IPv6 (SPMIPv6)—a novel anchors in software defined networking based Proxy Mobile IPv6 (ACM, scheme for mobility supported IP-WSNs. Sensors. 11(2), 1865–1887 (2011) New York, 2016), p. 106. http://doi.acm.org/10.1145/2857546.2857654 7. MM Islam, S-H Na, S-J Lee, E-N Huh, A novel scheme for PMIPv6 based 26. Y Kyung, Y Kim, K Hong, H Choi, M Joo, J Park, in 2016 IEEE Symposium on Wireless Sensor Network. Future Generation Information Technology: Second Computers and Communication (ISCC). Mobility-aware load distribution International Conference, FGIT 2010, Jeju Island, Korea, December 13-15, scheme for scalable SDN-based mobile networks (IEEE, Messina, 2016), 2010. Proceedings. (Springer, Berlin, 2010), pp. 429–438 pp. 119–124. https://doi.org/10.1109/ISCC.2016.7543725 8. MM Islam, TD Nguyen, AA Al Saffar, S-H Na, E-N Huh, in Computational 27. K Papagiannaki, N Taft, ZL Zhang, C Diot, in INFOCOM 2003. Twenty-Second Collective Intelligence. Technologies and Applications: Second International Annual Joint Conference of the IEEE Computer and Communications. IEEE Conference, ICCCI 2010, Kaohsiung, Taiwan, November 10-12, 2010. Societies. Long-term forecasting of Internet backbone traffic: observations Proceedings, Part III, ed. by J-S Pan, S-M Chen, and NT Nguyen. Energy and initial models, vol. 2 (IEEE, San Francisco, 2003), pp. 1178–11882. efficient framework for mobility supported smart IP-WSN (Springer, Berlin, https://doi.org/10.1109/INFCOM.2003.1208954 Heidelberg, 2010), pp. 282–291. http://dx.doi.org/10.1007/978-3-642- 28. EKRELHKNKSV Devarapalli, J Laganier, Heartbeat mechanism for proxy 16696-9_31 mobile IPv6. Technical report, RFC 5847, IETF Network Working Group 9. H-N Nguyen, C Bonnet, in 2008 5th IEEE International Conference on Mobile (June 2010). https://www.rfc-editor.org/info/rfc5847 Ad Hoc and Sensor Systems. Proxy mobile ipv6 for cluster based 29. R Jain, The art of computer systems performance analysis: techniques for heterogeneous wireless mesh networks (IEEE, Atlanta, 2008). https://doi. experimental design, measurement, simulation, and modeling. (Wiley, New org/10.1109/MAHSS.2008.4660097 York, 1991) 10. AJ Jabir, S Shamala, Z Zuriati, N Hamid, A comprehensive survey of the 30. S Deering, R Hinden, Internet Protocol, Version 6 (IPv6) Specification. current trends and extensions for the Proxy Mobile IPv6 Protocol. IEEE Technical report, IETF, RFC 2460 (1998). https://www.rfc-editor.org/info/ Syst. J. PP(99), 1–17 (2015) rfc2460 11. CKKRPBH Yokota, F Xia, Fast Handovers for Proxy Mobile IPv6. Technical 31. M Ghaleb, S Subramaniam, M Othman, Z Zukarnain, Predetermined path report, RFC 5949 (2010). https://www.rfc-editor.org/info/rfc5949 of mobile data gathering in wireless sensor networks based on network 12. MS Kim, S Lee, D Cypher, N Golmie, in Global Telecommunications layout. EURASIP J. Wirel. Commun. Netw. 2014(1), 51 (2014) Conference (GLOBECOM 2010). Fast Handover Latency Analysis in Proxy 32. T Issariyakul, E Hossain, Introduction to Network Simulator 2 (NS2). Mobile IPv6 (IEEE, 2010), pp. 1–5 (Springer, Boston, 2012), pp. 21–40 13. M-S Kim, S Lee, Load balancing and its performance evaluation for layer 3 33. K Fall, K Varadhan, The ns manual (formerly ns notes and documentation). and IEEE 802.21 frameworks in PMIPv6-based wireless networks. Wirel. The VINT project. 47, 19–231 (2005) Commun. Mob. Comput. 10(11), 1431–1443 (2010) 14. K Mun-Suk, L SuKyoung, A novel load balancing scheme for PMIPv6-based wireless networks. {AEU} - Int. J. Electron. Commun. 64(6), 579–583 (2010) 15. MS Kim, S Lee, in 2009 IEEE 20th International Symposium on Personal, Indoor and Mobile Radio Communications. Load balancing based on layer 3 and ieee 802.21 frameworks in pmipv6 networks (IEEE, Tokyo, 2009), pp. 788–792. https://doi.org/10.1109/PIMRC.2009.5450177 16. H Kong, Y Jang, H Choo, in Computational Science and Its Applications (ICCSA), 2010 International Conference On. An efficient load balancing of Mobile Access Gateways in Proxy Mobile IPv6 Domains (IEEE, Fukuoka, 2010), pp. 289–292. https://doi.org/10.1109/ICCSA.2010.67 17. J Dimple, C Kailash, A load reduction and balancing scheme for MAG operating in PMIPv6 domain. (Springer, Berlin, Heidelberg, 2013), pp. 128–136. http://dx.doi.org/10.1007/978-3-642-35864-7_19 18. S Qutub, T Anjali, in Electro/Information Technology (EIT), 2012 IEEE International Conference On. Load sharing mechanism for Mobile Access Gateways in PMIPv6 (IEEE, Indianapolis, 2012). https://doi.org/10.1109/EIT. 2012.6220760 19. F Teraoka, T Arita, in 2011 Third International Conference on Ubiquitous and Future Networks (ICUFN). PNEMO: a network-based localized mobility management protocol for mobile networks (IEEE, Dalian, 2011), pp. 168–173. https://doi.org/10.1109/ICUFN.2011.5949156 20. H Jiang, Load sharing support for MAGs in Proxy Mobile IPv6. Technical report, IETF-Draft (expired) (December, 2011) 21. T-T Nguyen, C Bonnet, Considerations of IP multicast for load balancing in Proxy Mobile IPv6 networks. Comput. Netw. 72(Supplement C), 113–126 (2014) 22. TT Nguyen, C Bonnet, in 2014 International Conference on Computing, Networking and Communications (ICNC). Load balancing mechanism for

Journal

EURASIP Journal on Wireless Communications and NetworkingSpringer Journals

Published: May 30, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off