Y. Afek, E. Gafni, and A. Ros'en. The Slide Mechanism with Applications in Dynamic Networks. In Proc. 11th ACM Symp. on Principles of Distributed Computing, pages 35--46, August 1992.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....[5, 9, 16] Dynamic networks are fixed wired networks which share some characteristics of ad hoc networks, since failure and repair of nodes and links is unpredictable in both cases. Research on dynamic networks has focused on total ordering [13] end to end communication, and routing (e.g. [1, 2]) Existing distributed algorithms will run correctly on top of ad hoc routing protocols, since these protocols are designed to hide the dynamic nature of the network topology from higher layers in the protocol stack (see figure 1(a) Routing algorithms on ad hoc networks provide the ability to ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proc. of 11th Annual Symp. on Prin. of Dist. Computing, pages 35--46, 1992.

....12, 24, 30] Dynamic networks are xed wired networks that share some characteristics of ad hoc networks, since failure and repair of nodes and links is unpredictable in both cases. Research on dynamic networks has focused on total ordering [19] end to end communication, and routing (e.g. [1, 2]) The # mutual exclusion problem involves a group of # processes, each of which intermittently requires access to an identical resource or piece of code called the critical section (CS) At most #,1####, processes may be in the CS at any given time. Providing shared access to resources through ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proc. of 11th Annual Symp. on Prin. of Dist. Computing, pages 35-46, 1992.

....(979) 847 8578. Dynamic networks are xed wired networks that share some characteristics of ad hoc networks, since failure and repair of nodes and links is unpredictable in both cases. Research on dynamic networks has focused on total ordering [17] end to end communication, and routing (e.g. [1,2]) Existing distributed algorithms will run correctly on top of ad hoc routing protocols, since these protocols are designed to hide the dynamic nature of the network topology from higher layers in the protocol stack (see Figure 1(a) Routing algorithms on ad hoc networks provide the ability to ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proc. of 11th Annual Symp. on Prin. of Dist. Computing, pages 35-46, 1992.

....[5, 9, 16] Dynamic networks are fixed wired networks which share some characteristics of ad hoc networks, since failure and repair of nodes and links is unpredictable in both cases. Research on dynamic networks has focused on total ordering [13] end to end communication, and routing (e.g. [1, 2]) Existing distributed algorithms will run correctly on top of ad hoc routing protocols, since these protocols are designed to hide the dynamic nature of the network topology from higher layers in the protocol stack (see figure 1(a) Routing algorithms on ad hoc networks provide the ability to ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proc. of 11th Annual Symp. on Prin. of Dist. Computing, pages 35--46, 1992.

....fr. 1 run a distributed pre processing algorithm to update the routing scheme and to make it adaptive to dynamic networks. It is willing that this maintaining algorithm has low message or time complexity. For more details on the dynamic case, we invite the reader to consult [1, 11] and [2, 5, 28] for endto end communication problems, where the goal is to guarantee communications between a xed pair of nodes in spite of link failures with the minimum memory space in the nodes and minimum communication messages. To illustrate the (static) routing problem, let us consider the following ....

Y. Afek, E. Gafni, and A. Ros en, The slide mechanism with applications in dynamic networks, in 11 th Annual ACM Symposium on Principles of Distributed Computing (PODC), ACM PRESS, Aug. 1992, pp. 35-46.

....maximum number of processors that must change their local states, upon a topology change from a legitimate state, so that the protocol is in a legitimate state. 1.2 Background and Motivation Many distributed protocols have been designed to cope with continuous dy 1. 2 1 namic changes [AAG87, AGH90, AM92, AGR92]. These protocols make certain assumptions about the behavior of processors and links during failure and recovery; for instance, most of them do not consider the possibility of processor crashes, and they assume that every corrupted message is identified and discarded. If failures are frequent, ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proceedings of the 11th Annual ACM Symposium on Principles of Distributed Computing, pages 35--46, New York, 1992. ACM.

....dynamic or asynchronous networks. In general, work on dynamic and asynchronous networks has been limited. In work related to load balancing for instance, an end to end communication problem, namely one in which messages are routed from a single source to a single destination, has been studied in [1, 7] on dynamic networks. Our scenario is substantially more involved since we are required to move load between several sources and destinations simultaneously. Another result on dynamic networks is the recent analysis of a local algorithm for the approximate multicommodity flow problem [5, 6] While ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proceedings of the 11th ACM Symposium on Principles of Distributed Computing, pages 35--46, August 1992.

.... frequent path stability: Infinitely often there is a period of time during which links forming at least one path between the sender and the receiver are operating (e.g. AGH90, H92] eventual connectivity: The only assumption is that there is no permanent sender receiver link cut (e.g. [AG88, AMS89, AG91, AGR92]) Almost all existing end to end protocols depend on having physical links that are, or can be made to be, well behaved in that the sequence of messages delivered is always a prefix of the sequence sent, i.e. no messages are lost in the middle. If processors do not crash, then this behavior ....

....O(nP(1 L=T ) where n is the number of processors in the system, and T is the retransmission parameter. Roughly speaking the time and message complexity are measured only for runs in which the time for a message to travel over a viable link is 1; this is comparable to the complexity measures in [AGR92] (which assumes a reliable data link layer) The number of possible paths P between two processors in the system is theoretically exponential in jEj, the number of links in the system. This could be a drawback for our protocol. However, most practical communication schemes 3 In practice, there is ....

[Article contains additional citation context not shown here]

Y. Afek, E. Gafni and A. Rosen. The Slide Mechanism with Applications in Dynamic Networks. In Proc. of the 11th Annual ACM Symposium on Principles of Distributed Computing, pp. 35--46, 1992.

....[8,12,21,26] Dynamic networks are xed wired networks that share some characteristics of ad hoc networks, since failure and repair of nodes and links is unpredictable in both cases. Research on dynamic networks has focused on total ordering [17] end to end communication, and routing (e.g. [1,2]) Existing distributed algorithms will run correctly on top of ad hoc routing protocols, since these protocols are designed to hide the dynamic nature of the network topology from higher layers in the protocol stack (see Figure 1(a) Routing algorithms on ad hoc networks provide the ability to ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proc. of 11th Annual Symp. on Prin. of Dist. Computing, pages 35-46, 1992.

....measure is the maximum number of processors that must change their local states, upon a topology change from a legitimate state, so that the protocol is in a legitimate state. 1. 2 Background and Motivation Many distributed protocols have been designed to cope with continuous dynamic changes (e.g. [AAG87, AGH90, AM92, AGR92]) These protocols make certain assumptions about the behavior of processors and links during failure and recovery; for instance, most of those works do not consider the possibility of processor crashes 2 and they assume that every corrupted message is identified and discarded. If failures are ....

Y. Afek, E. Gafni and A. Rosen, "The Slide Mechanism with Applications in Dynamic Networks," Proc. of the 11th ACM Symp. on Principles of Distributed Computing, pp. 35--46, 1992.

....dynamic or asynchronous networks. In general, work on dynamic and asynchronous networks has been limited. In work related to load balancing for instance, an end to end communication problem, namely one in which messages are routed from a single source to a single destination, has been studied in [1, 7] on dynamic networks. Our scenario is substantially more involved since we are required to move load between several sources and destinations simultaneously. Another result on dynamic networks is the recent analysis of a local algorithm for the approximate multicommodity flow problem [5, 6] While ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proceedings of the 11th ACM Symposium on Principles of Distributed Computing, pages 35--46, August 1992.

....dynamic or asynchronous networks. In general, work on dynamic and asynchronous networks has been limited. In work related to load balancing for instance, an end to end communication problem, namely one in which messages are routed from a single source to a single destination, has been studied in [1, 7] on dynamic networks. Our scenario is substantially more involved since we are required to move load between several sources and destinations simultaneously. Another result on dynamic networks is the recent analysis of a local algorithm for the approximate multicommodity flow problem [5, 6] While ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proceedings of the 11th ACM Symposium on Principles of Distributed Computing, pages 35--46, August 1992.

....measure is the maximum number of processors that must change their local states, upon a topology change from a legitimate state, so that the protocol is in a legitimate state. Background and Motivation Many distributed protocols have been designed to cope with continuous dynamic changes [AAG87, AGH90, AM92, AGR92]. These protocols make certain assumptions about the behavior of processors and links during failure and recovery; for instance, most of them do not consider the possibility of processor crashes and they assume that every corrupted message is identified and discarded. If failures are frequent, ....

Y. Afek, E. Gafni and A. Rosen, "The Slide Mechanism with Applications in Dynamic Networks," Proc. of the 11th ACM Symp. on Principles of Distributed Computing, pp. 35--46, 1992.

....dynamic or asynchronous networks. In general, work on dynamic and asynchronous networks has been limited. In work related to load balancing for instance, an end to end communication problem, namely one in which messages are routed from a single source to a single destination, has been studied in [1, 7] on dynamic networks. Our scenario is substantially more involved since we are required to move load between several sources and destinations simultaneously. Another result on dynamic networks is the recent analysis of a local algorithm for the approximate multicommodity flow problem [5, 6] While ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proceedings of the 11th ACM Symposium on Principles of Distributed Computing, pages 35--46, August 1992.

No context found.

Y. Afek, E. Gafni, and A. Ros'en. The Slide Mechanism with Applications in Dynamic Networks. In Proc. 11th ACM Symp. on Principles of Distributed Computing, pages 35--46, August 1992.

No context found.

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proc. 11th ACM Symp. on Principles of Distributed Computing, pages 35--46, August 1992.

....i.e. each pair of neighboring nodes share a pair of atomic read write registers. Each atomic operation is either a read, or a write, or an internal operation (in this respect we follow [DIM94] The algorithms presented here can be adapted to the message passing model by using the techniques of [AB89, AGR92, AV91, Var92]. This transformation to the message passing model is possible under the realistic assumptions that (1) the capacity of each link is bounded, and (2) for each link one of its incident nodes is distinguished as the link master (an assumption that trivially holds if nodes have unique identities) ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proc. 11th ACM Symp. on Principles of Distributed Computing, pages 35-46, August 1992.

....paper, a randomized election algorithm for anonymous networks was developed by S. Dolev, Israeli and Moran [DIM91] The algorithm and its extensions are presented in the shared memory model. They can be adapted to run in a message passing communication network by using the techniques of [AB89, AGR92, AV91] This transformation to the message passing model is possible under the realistic assumptions that (1) the capacity of each link is bounded, and (2) for each link one of its incident nodes is distinguished as the link master (an assumption that holds trivially if nodes have unique ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proc. 11th ACM Symp. on Principles of Distributed Computing, pages 35--46, August 1992. 41

....each pair of neighboring nodes share a pair of atomic read write registers. Each atomic operation is either a read, or a write, or an internal operation (in this respect we follow [DIM94] The algorithms presented here can be adapted to the message passing model by using the techniques of [AB89, AGR92, AV91, Var92] This transformation to the message passing model is possible under the realistic assumptions that (1) the capacity of each link is bounded, and (2) for each link one of its incident nodes is distinguished as the link master (an assumption that trivially holds if nodes have unique ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proc. 11th ACM Symp. on Principles of Distributed Computing, pages 35--46, August 1992.

....the network with each item [Vis83, AE86] However, this solution has the drawback that the amount of communication needed per data item grows unboundedly as the number of messages grows. Recently, the study of end to end protocols with bounded communication complexity received a lot of attention [AG88, AMS89, APV91, AG91, AGR92]. In this paper, we concentrate on bounded (in fact, polynomial) communication complexity protocols. space complexity: Another important complexity measure is the space complexity the amount of space needed at the nodes, per incident link. Notice that the unbounded sequence numbers ....

....(total number of bits) Space Complexity (bits per incident link) Vis83, AE86] unbounded: 1 unbounded: 1 [AG88] Alg. 1. unbounded: 1 constant : O(D) AG88] Alg. 2. exponential: O(D Delta exp(n) logarithmic: O(log n D) AMS89] polynomial : O(n 9 mD) polynomial : O(n 5 D) [AGR92] polynomial : O(n 2 mD) linear : O(nD) AG91] polynomial : O(nm log n mD) linear : O(n D) PRESENT WORK: polynomial : O(n 2 mD) logarithmic : O(log n D) Our Techniques and Previous Work: The starting point of our investigation is the Slide protocol of [AGR92] and the Majority algorithm ....

[Article contains additional citation context not shown here]

Y. Afek, E. Gafni, and A. Ros'en. The slide mechanism with applications in dynamic networks. In Proc. 11th ACM Symp. on Principles of Distributed Computing, pages 35--46, August 1992. 10

....O(n) amortized message complexity end to end protocol. Preliminary versions of the various results in this paper appeared in Proc. of the 30th IEEE Annual Symp. on Foundation of Computer Science 1989, and Proc. of the Eleventh Annual ACM Symp. on Principles of Distributed Computing, 1992 [AMS89, AGR92] y Department of Computer Science, Tel Aviv University, Tel Aviv 69978, Israel. z Computer Science Department, Johns Hopkins Univ. and Dept. of Mathematics and Lab. for Computer Science, M.I.T. Cambridge, MA 02139; Supported by Air Force Contract TNDGAFOSR 86 0078, ARO contract DAAL0386 ....

....of the IDA [Rab89] the space complexity of the sender and the receiver is O(n 2 m 2 log n) 2 5 Conclusion This paper introduces the slide protocol and uses it to provide the first polynomial complexity end to end communication protocol in dynamic networks. Since its initial publication [AGR92] slide has been used as the basis for several new algorithms, including the elegant self stabilizing protocols [AV91, APSV91, Var92] a load balancing scheme [AAMR93] and a multi commodity flow algorithms [AL94, AL93] We believe it will find further applications in network protocol design as ....

Y. Afek, E. Gafni, and A. Rosen. The slide mechanism with applications in dynamic networks. In Proc. 11th ACM Symp. on Principles of Distributed Computing, pages 35--46, August 1992.

....above reset and self stabilizing solutions work only if faults are not too frequent, as they require that a message is transmitted over the computed path. Communication During Frequent Faults: We consider a setting where failures occur frequently. The so called end to end communication problem [AG88, AMS89, AGR92] is to deliver, in finite time, data items from a sender processor to a receiver processor, where data items are given in an on line fashion to the sender, and must be output in the same order, without duplication or omission at the receiver processor. This should be done even if there does not ....

....drawback that the message size increases unboundedly with the number of items sent; hence, the amount of communication needed per data item grows unboundedly with the number of data items. Recently, the study of end to end protocols with bounded communication complexity received a lot of attention [AG88, AMS89, APV91, AG91, AGR92]. In this paper, we concentrate on bounded (in fact, polynomial) communication complexity protocols. Space Complexity: Another important complexity measure is the space complexity the amount of space needed at the processors, per incident link. Notice that the unbounded sequence numbers ....

[Article contains additional citation context not shown here]

Y. Afek, E. Gafni, and A. Ros'en. The Slide Mechanism with Applications in Dynamic Networks. In Proc. 11th ACM Symp. on Principles of Distributed Computing, pages 35--46, August 1992.

No context found.

Y. Afek, E. Gafni, and A. Ros'en. The Slide Mechanism with Applications in Dynamic Networks. In Proc. 11th ACM Symp. on Principles of Distributed Computing, pages 35--46, August 1992.

No context found.

Y. Afek, E. Gafni, and A. Ros en, The slide mechanism with applications in dynamic networks, in 11 Annual ACM Symposium on Principles of Distributed Computing (PODC), ACM PRESS, Aug. 1992, pp. 35-46.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC