journal article
LitStream Collection
Inouye, Jon; Konuru, Ravindranath; Walpole, Jonathan; Sears, Bart
doi: 10.1145/142854.142859pmid: N/A
Recent times have witnessed rapid advances in microprocessor technology resulting in an order of magnitude performance improvement every few years. These developments in hardware have been paralleled by several prominent trends in operating system design, the most notable being a move towards message-passing micro-kernels. However, operating system performance has not kept pace with that of the underlying hardware. It has become apparent that design changes to enhance processor performance can have adverse effects on operating system performance. This problem arises when the architectural assumptions implicit in an operating system's design are inappropriate for the architectures on which it executes.This paper examines one specific area in which operating system design assumptions appear to be in conflict with trends in modern processor architecture. We focus on the performance effects of virtually addressed caches on two contemporary operating systems (Mach and Chorus). We present experimental results to illustrate the impact of virtually addressed caches on the performance of primitive virtual memory operations, and higher-level operations, such as inter-process communication, that utilize these primitive operations. The main goal of the paper is to encourage operating system designers to revisit some of the basic architectural assumptions implicit in modern operating system designs.
doi: 10.1145/142854.142865pmid: N/A
This paper develops an abstract model which presents a method of uniform description of different rollback recovery control algorithms for distributed systems. We first developed a general definition of the distributed rollback recovery control problem. The concept of a distributed recovery control system ( DRC system ), consisting of distributed recovery control units ( DRC units ), is proposed to model recovery with various control granularities. Then, we developed a graph model, called dependency graph , for distributed rollback recovery control algorithms. An atomic subgraph is defined as a subgraph induced by a set of nodes which has no outgoing arcs to other nodes in the graph. Committing and aborting atomic actions can be modeled as identifying atomic subgraphs. Next, we defined two kinds of dependency graphs: checkpoint graphs and unit graphs , based on the dependency relation defined by rollback propagation. We have shown that various types of distributed recovery control algorithms can be classified based on the identifications of atomic subgraphs in these two graphs. Therefore, using the model may allow us to describe existing algorithms in a uniform way and, more importantly, to find new algorithms.
doi: 10.1145/142854.142870pmid: N/A
Multicast protocols must often provide a given property on the order in which messages are delivered. This delivery order may be consistent with local dependences between messages or with causal dependences. A local delivery is defined according to the emitting order observed at the same process source. A causal delivery is defined according to the causal relations observed between different emitting events. If an arbitrary delivery order is applied (when there are no particular dependences) the multicast is said to be totally ordered. In this paper we study totally and causally ordered multicast protocols i.e. multicast protocols that guarantee a total order conforming with the causal dependences. We propose a new method that begins to build a total order to obtain a causal and total multicast, when all current methods propose to build first a causal order then a total order (developed using the services of this causal layer).This new approach results from a formal study of ordering properties. In the first main result we show that a broadcast that guarantees a local and total order also provides a causal broadcast. Then we show that a multicast protocol that guarantees a local and a total order with additional constraints is also a causal and total multicast. This result makes no assumption about structure of overlapping groups. We finally develop the scheme of a multicast protocol using a centralized server of order. This solution is efficient in terms of messages and delay for a small number of sites. It allows all types of messages (unicast, multicast, broadcast) to mix while getting a causal and total ordering.
Kehne, A.; Schönwälder, J.; Langendörfer, H.
doi: 10.1145/142854.142872pmid: N/A
The Kerberos authentication service, a part of MIT's Project Athena, is based on the Needham and Schroeder protocol. Timestamps depending on reliable synchronized clocks are used to guarantee the freshness of messages. As an improvement, we present a nonce-based protocol offering the same features as Kerberos. We generate a ticket in an initial message exchange which includes a generalized timestamp. Checking this generalized timestamp is left to the principal who created it. Consequently we do not need synchronized clocks. Our protocol has the property of using a minimal number of messages to establish an authenticated session key.
Showing 1 to 8 of 8 Articles