Enabling Efficient Content Location and Retrieval in Peer-to-Peer Systems by Exploiting Locality in Interests K u n w a d e e Sripanidkulchai Bruce Maggs Hui Zhang Carnegie M e l l o n U n i v e r s i t y { k u n w a d e e , b m m , h z h a n g ) @cs.cm u .ed u Services on the Interact are evolving from centralized client-server architectures to fully distributed architectures. End-hosts are becoming more ubiquitous, more powerful, and more involved in providing services. The wide-spread adoption of Interact access as a utility sexvicc is enabling new modes of interaction between end-hosts. End-hosts can provide services as well as use services. We call systems based on such service architectures peer-to-peer systems, and end-hosts participating in such systems peers. Our interests lie Jn peer-to-peer content publishing and distribution, where peers publish content to the system and download content from the system. Peer~ contribute storage and collaborate while participating in the system. Downloading content involves locating peers who have copies of the content, selecting st peer, and retrieving a copy from that peer. The characteristics unique to peer-to-peer systems are dynamicit), and variability. For example, content in the system is dynamically replicated, and peers dynamically join and leave the system. Furthermore, peers have a wide range of network access speeds, and variability in load and stvailable bandwidth at each peer can be extensive. To study variability in performance, we measured ping times to endhosts on the Intmmet at 30-second intervals over a 24-hour period. Variability in ping time implies variability in download performance. We collected IP addresses of peers participating in Gnutellst [1], a filesharing application, on April 16, 2001. Out of the 58,400 addresses collected, 24.54 were randomly chosen and pinged on April 23 and May I, 2001. Figure 1 depicts the measured ping time to a peer with cable modem access. The ping times vary over st wide range from 300 milfiseconds to 24 seconds. The standard deviation is on the order of seconds, which is typical for a third of the peers measured in our experiments. Unlike servers, end-hosts are not exclusively provisioned foe providing service. End-hosts can be used to run many applications locally while actively participating in peer-to-peer content distribution. F o r m a n y hosts, bandwidth is a scarce resource. Supporting st f e w concurrent downloads is feasible. But, additional connections can significantly degrade d o w n l o a d performance. Protocols designed for peer-topeer systems need to takz into account its d y n a m i c and variable nature. - (a) Peer list overlay (b) Locality in ime~re,mmlatioNhlp 2. Usiag Iocalityinintemata. - "....... T j. ~oo lilmulmJlon I m a m (s) t Zt ,~o moo ~ loooo l m t o ~oo . il ~,j~m.,~) . .."1" loooo l ~ o (a) M i . rate (b) Peer lilt ai~ l r i . 3. ped'ormace of asing lecallty in iatereatsto lecate coatmL There are many challenges in designing peer-to-peer content distriImfion systems. In this work, we address the challenge of locating and retrieving content in st scalable, efficient, and distributed w a y when peers and the network have extremely high variability in performance. Existhag solutions, such as Tapestry [6], Chord [5], CAN [3], and P u n y [4] have addressed scelability. However, no solution explicitly addresses performance. In order to achieve good performance, it is necessary to consider dynamic conditions. Incorporating dynamic performance into existing protocols is not trivial because it can greatly reduce scalshility. We propose a novel solution based on locality in interests to identify a small set of ~ fur which to maintain dynamic performance state. Peers self-organize into groups. Each peer maintains a list of peers who share similar inmmests. Peers on the list are ranked based on current inJ ~i " Mlwl.~I tmests and dy,amic performance. Content is located by querying peers on one's list. Figure 2(a) depicts a peer list overlay constructed on top of Gnutella. When content cannot be found through the list, peers use an underlying location mechanism, such as flooding in Gnutella or lookups in Chord, to locate content. In our initial evaluation, we use the following heuristic to identify locality in interests: peers that have the content we are looking for share the same interests. Figure 2('o) illustrates this relationship. The peer in the middle is looking for content A, B, and C, which can all be found at the peer at the far left. To evaluate the benefits of using locality in interests to locate content, we run simulations using the Boeing corpo~te web proxy traces [2] to drive the request stream. We compare three content location algorithms: ask random peers, ask peers who share the same interests (l-hop), and ask peers and peers of peers with the same interests (2-hops). The average, maximum, and minimum miss rate observed from 16 simulation using all three algorithms is shown in Figure 3(a). The miss rate is defined as the percentage of requests for which content that already exists in the system cannot be found. Using the random algorithm results in a 35% miss rate. The miss rate using locality in inte=ests is significantly lower: 10% whea asking peers 1 hop on the peer list and down to 5% when ask peers'2 hops on the list. Figure 3(b) depicts the average size of the peer list maintained at each node. On average, m a i n t s t i n l n g st list of 8 peers provides sufficiently low miss rates. Wc demonstram that locating content among peers with shared interests is effective and incurs low overhead. We are currently exploring heuristics to refine our solution by ranking peers in the list based on dynamic performance and boostrappiug the list using alternative mechanisms. We are also implementing our solution for Gnutella. Please visit our webpage, http://www.cs.cmu.edu/'kunwadee/research/p2p, for more information about our research and for the implementation we plan to release shortly. REFERBNCES ~] OImtcllm bl~p'JlguulrJItweBe~om. " ]Mherch1999 J'MeJd°wl'-Bolingpfmly[°sn" &vmi -lal~atflp:llmsemehmalO"~'vLedtt/PUbf°°ei~'" " I~t S. aamaimmy, P. F n m d t , ~ L Hnadley, st. lTJerp, aad S. Shmk~z. :...: i;...,.: wmlr= Pme~rJmf# t~fACM PC/OCOMM, Auguat 2001. . . [4] A. ltownllnm aac[p. Dm,ek,.d, PmltrT: Sga.lsble.,di~lnbe~d object Inflttbm md t~lUnJI for Imse-Kale [5] I. smir.~lLMo~s~D. Kaq~r~M.P. lCJmslmek, m d H . Balakrhttumm. ~ A _m~_blep:~-m-pemr Iookap re'vim for Intm~m ~ w . a t l n M . Pmceeel#~s ~f AC.M ~GCOMM, Ausmm ~ ] . [6] B . ~ h ~ , J.l~lblatowiez, mlM]A-Jo~'ph. 'ral~JUPj: A n i m f m t r u c l u m f o r ~ f m d t - t d m ' n a l l J r . a tton and muting. U. C. lherlmlwyTecAnJml l~epon UC~LSD-OI-I l q i , AIHI 7J)Ol. A .~¢allable mm~t-,mdmmaele - ~-p~ Wm:mL ~ j n ~ J b r ~ , b ~ o n . i% !. Pinglimem an end-I~ with c~,b~modem~ a . s. ACM SIGCOMM 80 Computer Communication Review

/lp/association-for-computing-machinery/enabling-efficient-content-location-and-retrieval-in-peer-to-peer-7O3myj8m2z