Y. Afek, H. Attiya, A. Fekete, M. J. Fischer, N. Lynch, Y. Mansour, D. Wang, L. D. Zuck. Reliable Communication over Unreliable Channel. J. of the ACM 40(5):1087-1107, 1993.  Home/Search   Document Details and Download   Summary   ACM   TOC   Related Articles   Check

....are already considered in the original paper on the SW protocol by Stenning [20] The protocols for such chan nels (called transport protocols) such as TCP, can transmit data over very large networks such as Internet. Note that not for all types of transport channels a SW protocol exists. As [2] shows, for a fully asynchronous system and channels that can both lose and reorder messages, it is impossible to design an efficient transmission protocol that uses bounded sequence numbers. Similar result is proved for systems that can both reorder and duplicate messages [23] In [18] unbounded ....

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6):1267--1297, 1994.

....In the deterministic setting, i.e. the context of this paper, solutions for the stream of messages end to end problem di er according to the type of faults. Wand and Zuck [38] have shown that any protocol tolerating both packet reordering and duplication requires unbounded headers. Afek et al. [6] have shown that packet reordering and loss create the same e ect, that is either unbounded headers or non termination (i.e. the same packet can be received an unbounded number of times) Fekete and Lynch [21] have shown that just packet loss implies that some header information must be attached ....

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D.-W. Wang, and L. Zuck. Reliable Communication Over Unreliable Channels. Journal of the ACM 41(6):1267-1297 (1994).

....environments where reordering problems do not occur, this does not present a problem. Afek and others have shown, howevex , that protocols using only a finite message alphabet can never solve the sequence transmission problem in environments where both reordering and deletion errors may occur, see [1]. Thus it is essential that our knowledge based protocol for TCP uses an infinite message alphabet or messages growing in size. In the previous section we discussed Halpern and Zuck s result about the eventual attainment of depth 4 knowledge when using protocol B. It is surprising to note that ....

Y. Afek, H. Attiya, A. Fekete, M. J. Fischer, N. Lynch, Y. Mansour, D. Wang and L. D. Zuck, Reliable communication over unreliable channels, Journal of the ACM, Vol. 41, No. 6 (1994), pp. 1267-1297.

....environments where reordering problems do not occur, this does not present a problem. Afek and others have shown, however, that protocols using only a nite message alphabet can never solve the sequence transmission problem in environments where both reordering and deletion errors may occur, see [1]. Thus it is essential that our knowledge based protocol for TCP uses an in nite message alphabet or messages growing in size. In the previous section we discussed Halpern and Zuck s result about the eventual attainment of depth 4 knowledge when using protocol B. It is surprising to note that ....

Y. Afek, H. Attiya, A. Fekete, M. J. Fischer, N. Lynch, Y. Mansour, D. Wang and L. D. Zuck, Reliable communication over unreliable channels, Journal of the ACM, Vol. 41, No. 6 (1994), pp. 1267-1297.

....environments where reordering problems do not occur, this does not present a problem. Afek and others have shown, however, that protocols using only a nite message alphabet can never solve the sequence transmission problem in environments where both reordering and deletion errors may occur, see [1]. Thus it is essential that our knowledge based protocol for TCP uses an in nite message alphabet (or messages growing in size) Accumulating knowledge The theorem below holds in communication media where there may be deletion and reordering errors, but no other kinds. In particular, there ....

Y. Afek, H. Attiya, A. Fekete, M. J. Fischer, N. Lynch, Y. Mansour, D. Wang and L. D. Zuck, Reliable communication over unreliable channels, Journal of the ACM, Vol. 41, No. 6 (1994), pp. 1267-1297.

....there has also been significant theoretical work in the study of reliable message delivery protocols. The earlier work in the area considered just the possibility of reliable message delivery and mostly in a purely asynchronous setting. This is the case for the impossibility results of Afek et al. [1] and Fekete, Lynch, Mansour, and Spinelli [6] In [2] Attiya, Dolev, and Welch attain further results for the asynchronous model based on the minimum amount of information that must be maintained between incarnations of a connection in the presence of crashes. None of these papers examines the ....

....our assumption that the protocol delivers 8 messages reliably. In the executions in the proof, the delay on some packets is 2d k. Since k may be some arbitrary value set by the server, the proof requires that there is no MPL. 4 Conclusion There has been significant theoretical results [1, 6, 2, 3, 10, 13] on the limitations of connection management and reliable message delivery protocols under various timing and failure assumptions. Our work adds a new dimension to traditionally studied problems by adding conditional requirements. In our work we formally define what we call the conditionally fast ....

Afek, Y., Attiya, H., Fekete, A., Fischer, M., Lynch, N., Mansour, Y., Wang, D.-W., and Zuck, L. Reliable communication over unreliable channels. Journal of the ACM 41, 6 (November 1994), 1267--1297.

....does not work if the physical channels can reorder mes48 sages. In fact, Lynch, Mansour, and Fekete [LMF88] showed that no protocol with bounded headers can work over non FIFO physical channels, if the best case number of packets required to deliver each message must be bounded. Attiya et al. [AFWZ89] complete the picture by showing that this latter assumption is necessary that there is a (not very practical) protocol using bounded headers if the best case number of packets required to deliver one message is permitted to grow without bound. Baratz and Siegel [BS88] developed link protocols ....

H. Attiya, M. Fischer, D. Wang, and L. Zuck. Reliable communication over an unreliable channel. In Proceedings of the Twentieth Annual ACM Symposium on Theory of Computing, Association for Computing Machinery, 1989.

....however, it will not be possible to prove that C implements A because the definition of implements requires that the inputs and outputs are identical, not just the externals. 3 An example verification In the sequel we specify and verify a communication protocol originally due to Attiya et al. [3]. This protocol achieves FIFO communication over channels which may lose and reorder messages, but may not duplicate them. Let us call such channels non duplicating (ND) The remarkable fact is that the protocol requires only a single header bit. Starting from [3] Nipkow verified a slightly ....

....originally due to Attiya et al. 3] This protocol achieves FIFO communication over channels which may lose and reorder messages, but may not duplicate them. Let us call such channels non duplicating (ND) The remarkable fact is that the protocol requires only a single header bit. Starting from [3], Nipkow verified a slightly optimized version of this protocol some 4 years ago, using Isabelle merely as a theorem prover for the hand generated verification conditions. To simplify matters he also assumed that the underlying channels do not lose messages. In the following we recast Nipkow s ....

[Article contains additional citation context not shown here]

Hagit Attiya, Alan Fekete, Michael Fischer, Nancy Lynch, Yishay Mansour, DaWei Wang, and Lenore Zuck. Reliable communication over unreliable channels. Draft version, 1990.

....general purpose theorem provers because model checking [4] already provides a successful automatic approach to the verification of finite state systems. IOA were chosen as the vehicle for our study because they have become popular for specifying and verifying distributed algorithms both on paper [8, 2] and with machine assistance [10, 6, 13, 7] The unique aspect of our work is the fact that we have formalized and verified the meta theory of IOA on top of which we carried out our case study. Thus IOA are objects in the logic, just like natural numbers or lists, which can be manipulated by ....

....verification conditions. To simplify matters he also assumed that the underlying channels do not lose messages. In the following we recast Nipkow s specification and verification in the IOA formalization presented above. In the mean time the above protocol has been simplified considerably [2] using the following divide and conquer approach: 1. FIFO communication can be implemented on top of order preserving (OP = duplication and loss, but no reordering) channels using standard protocols like the Alternating Bit. 9 2. OP communication can be implemented on top of ND channels using a ....

Yehuda Afek, Hagit Attiya, Alan Fekete, Michael Fischer, Nancy Lynch, Yishay Mansour, Da-Wei Wang, and Lenore Zuck. Reliable communication over unreliable channels. Journal of the ACM. To appear.

....message using only a constant number or O(1) of low level messages once the network starts behaving ideally, must use unbounded headers. Such protocols are called O(1) bounded protocols. A formal definition follows later. O(n) bounded protocols that use bounded headers have been shown to exist by [AAF ] and [TL90]. MS89] shows the necessity of an unbounded receive buffer to transmit n high level messages with less than n headers. It also shows that if packets are lost on the link with some probability q, then, with a high probability, there exists an exponential lower bound on the number of low level ....

Y Afek, H Attiya, A Fekete, M Fischer, N Lynch, Y Monsour, D Wang, and L Zuck. Reliable communication over unreliable channel. JACM, pages 0--0.

....repaired. It reflects the usefulness of the channel. Without such a property (e.g. with permanent network partitioning) any interesting distributed problem would be trivially impossible to solve. Properties 2 and 3 define a communication channel similar to the non duplication channel of [1] which is a natural abstraction of the service provided by a connectionless network layer. 2.2 Best effort communication channels Informally, best effort channels do their best to provide the abstraction of a reliable channel, but only as long as no process or link failure is suspected. A ....

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, and L. Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6), July 1994.

....can overcome an unbounded (but finite) number of omission failures. The logarithmic space requirement stems from our lack of assumption on message ordering. Note that there can exist no constant memory protocols providing reliable multicast over an unreliable network that can re order messages [1]. In our solution, the memory requirements increase logarithmically due to the use of monotonic counters. In any practical execution, 64 or 128 bits of memory will be more than enough. Therefore our solution is reasonable for practical use. Finally, we prove minimality of 3W(om) for solving ....

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, W. Dai-Wei, and L. Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6):1267--1297, November 1994.

....synchronized so that the information provider will not overwrite the information before the other process has observed the content. Both algorithms share the same idea in the dynamic adjustment of the Delta parameter. 6 A similar showdown situation has been addressed by Afek, et al. [1] in solving the sequence transmission problem in an unreliable packet switching network. The rest of the paper is organized as follows. Section 2 presents the multiparty interaction scheduling problem. The message passing solution is presented in Section 3, and the shared memory solution in ....

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. A. Lynch, Y. Mansour, D.-W. Wang, and L. Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6):1267--1297, November 1994.

....are necessary in order to manage incarnations, under various requirements and assumptions. LeLann and LeGoff [18] show that a connection cannot be established by protocols of a particular form. Other theoretical studies of communication protocols have mostly concentrated on the data link layer [1, 3, 13, 16, 20, 22, 26, 29]. Most of this work concerns implementing protocols using bounded size packets, an issue we do not address. Our protocols for message transfer on bounded capacity FIFO networks use an idea from self stabilizing protocols for cleaning the system with a new label. This idea was first employed in ....

....was not lost but only delayed in the non FIFO network. Thus we can steal a single packet from each replay and keep it in the network. Then we deliver the packets we have collected to the receiver, tricking it into acting as if the reference execution is executed. A similar technique is used in [1]. Theorem 2.1 There is no amnesic incarnation management protocol for non FIFO losing networks, for any loss(2) only NProblem predicates. Sketch of proof: Assume in contradiction there is such a protocol. The definition of a correct protocol implies that there exists a ping pong schedule fl ....

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D.-W. Wang and L. Zuck, Reliable Communication over Unreliable Channels, Technical Report YALEU/DCS/TR-853, Department of Computer Science, Yale University, October 1992. To appear in Journal of the ACM.

....significant bit of the sequence number. It tolerates both packet loss and duplication. With the possibility of packet loss, Fekete and Lynch [14] have proved that there are no correct headerless protocols; thus the alternating bit protocol uses the shortest possible headers. Counting protocols [2, 26] are another way to handle packet reordering and loss. They are based on the following observation: Suppose R knows an upper bound b on the number of old packets in the network at a particular time. Then R can ensure that it has a new packet by waiting for b 1 copies of the same packet to ....

....that it has a new packet by waiting for b 1 copies of the same packet to arrive. In these protocols, the value of b depends on the number of packet losses that have occurred. Hence, b and the number of packets successfully transmitted per message can be arbitrarily large. However, Afek et al. [2] have proved that any protocol tolerating both packet reordering and loss either requires unbounded sequence numbers or can cause R to receive an unbounded number of packets per message. 3 General Networks For general networks, end to end communication is not possible if there is a permanent cut ....

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. Zuck, Reliable Communication Over Unreliable Channels, JACM, vol. 41, no. 6, 1994, pages 1267-- 1297.

....If p i stb sends messages m ; m k Gamma1 to p j , and p i indefinitely delays stb sending any further message to p j , then p j eventually stbreceives messages m ; m k Gamma1 . Properties 2.1 and 3. 1 define a communication channel similar to the nonduplication channel of [1], which is a natural abstraction of the service provided by a connectionless network layer. Property 3.2 slightly constrains the unreliability of the communication channels. It basically says that, if a green process p i stb sends a sequence [m ; m k Gamma1 ] of messages to a green ....

....if m 2 M S fg. The unique configuration that results from applying s = p; m; d; A) to C = I; M) is noted s(C) A null step s = is applicable to any configuration C = I; M ) and s(C) C. A schedule of an algorithm A is a (possibly finite) sequence of steps of A and null steps, noted S = S[1]; S[2] S[k] A schedule S is applicable to a configuration C if (1) S is the empty schedule, or (2) S[1] is applicable to C, S[2] is applicable to S[1] C) etc. A partial run of A using a red detector D r , is a tuple Upsilon = F; R; C; S; T where, F is a failure pattern, R is ....

[Article contains additional citation context not shown here]

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, and L. Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6), July 1994.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. J. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. D. Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6):1267--1297, 1994.

....of the other theoretical work in the area of reliable message delivery has considered somewhat different problems in different settings. Afek et al. and Fekete, Lynch, Mansour, and Spinelli prove impossibility results for different types of reliable communication in a purely asynchronous setting [2, 10]. In [3] Attiya, Dolev, and Welch attain further results for the asynchronous model based on the minimum amount of information that must be maintained between connections in the presence of crashes or between active incarnations of a crashes. None of these papers examines the amount of time or ....

Yehuda Afek, Hagit Attiya, Alan Fekete, Michael Fischer, Nancy Lynch, Yishay Mansour, Da-Wei Wang, and Lenore Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6):1267--1297, November 1994.

....additional operators that are used for timed automata are hiding and renaming. The hiding operator is used to reclassify output actions to be internal actions so that they are no longer visible to the environment. The renaming operator renames some subset of the actions of a timed automaton. See [4, 1] for a formal discussion of the hiding and renaming operators for timed automata. 2.2 Clock Automaton Model As a special case of a timed automaton we define a clock automaton. A clock automaton is a timed automaton with a state component, clock, which represents a clock time. Definition 2.3 ....

....the now state component s . The messages sent over edges are taken from an arbitrary message set M. In order to simplify the proofs we assume that each message sent is unique, i.e. the same message cannot be sent twice in a given execution. It is easy, though tedious, to remove this restriction [1]. In the following sections we describe how each component of a distributed system is modeled by timed automata. 3.1 Algorithm The timed automaton modeling node v, is referred to as A, Since the timed automata A, encode the algorithm, they must be supplied and proved to be correct by the ....

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. Zuck, Reliable Communica- tion over an Unreliable Channel, LCS TM-447, October, 1992. Also submitted to JACM.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. J. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. D. Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6):1267--1297, 1994.

....above. Harvey and Lynch [7] propose a number of the problems considered here, and obtain some of the initial results on duplication. In the fully asynchronous setting, Fekete, Lynch, Mansour, and Spinelli, and Afek et al. prove impossibility results for different types of reliable communication [5, 1]. Wang and Zuck [17] consider the sequence transmission problem, in which a sender must transmit a specified sequence of data to a receiver over a faulty channel. Further results in an asynchronous model, based on the minimum amount of information that must be maintained between connections, are ....

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, L. Zuck, "Reliable communication over an unreliable channel," to appear in Journal of the ACM. (Also Technical Memo MIT/LCS/TM-447.)

....above. Harvey and Lynch [7] propose a number of the problems considered here, and obtain some of the initial results on duplication. In the fully asynchronous setting, Fekete, Lynch, Mansour, and Spinelli, and Afek et al. prove impossibility results for different types of reliable communication [5, 1]. Wang and Zuck [17] consider the sequence transmission problem, in which a sender must transmit a specified sequence of data to a receiver over a faulty channel. Further results in an asynchronous model, based on the minimum amount of information that must be maintained between connections, are ....

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, L. Zuck, "Reliable communication over an unreliable channel," to appear in Journal of the ACM. (Also Technical Memo MIT/LCS/TM-447.)

No context found.

Y. Afek, H. Attiya, A. Fekete, M. J. Fischer, N. Lynch, Y. Mansour, D. Wang, L. D. Zuck. Reliable Communication over Unreliable Channel. J. of the ACM 40(5):1087-1107, 1993.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. J. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. D. Zuck. Reliable Communication over Unreliable Channel. J. of the ACM 40(5):10871107, 1993.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6):1267--1297, 1994.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D.-W. Wang, and L. Zuck. Reliable Communication Over Unreliable Channels. Journal of the ACM 41(6):1267-1297 (1994).

No context found.

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D.-W. Wang, and L. Zuck. Reliable communication over unreliable channels. J. ACM, 41(6):1267--1297, Nov. 1994.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D.-W. Wang, and L. Zuck. Reliable Communication Over Unreliable Channels. Journal of the ACM 41(6):1267-1297 (1994).

No context found.

Yehuda Afek, Hagit Attiya, Alan Fekete, Michael Fischer, Nancy Lynch, Yishay Mansour, Dai-Wei Wang, and Lenore Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6), pages 1267--1297, November 1994.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41:1267-1297, 1994.

No context found.

Afek, Attiya, Fekete, Fischer, Lynch, Mansour, Wang, and Zuck, Reliable Communication Over Unreliable Channels, JACM vol. 41, no. 6, 1994, pages 1267--1297.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. J. Fischer, N. Lynch, Y. Mansour, D-W. Wang, and L. D. Zuck. Reliable communication over unreliable channels. Technical Report YALEU/DCS/TR-853, Yale University, 1991.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. Zuck. Reliable communication over unreliable channels.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. Zuck. Reliable Communication Over Unreliable Channels. Technical Report YALEU/DCS/TR-853, Department of Computer Science, Yale University, October 1992.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. J. Fischer, N. Lynch, Y. Mansour, D. Wang, L. D. Zuck. Reliable Communication Over Unreliable Channel. Manuscript. 1990.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. Zuck, Reliable Communication Over Unreliable Channels, Journal of the ACM, 41(6), (1994), 1267-1297.

No context found.

Afek, Attiya, Fekete, Fischer, Lynch, Mansour, Wang, and Zuck, Reliable Communication Over Unreliable Channels, JACM, vol. 41, no. 6, 1994, pages 1267-1297.

No context found.

Y. Afek, H. Attiya, A. Fekete, M. J. Fischer, N. Lynch, Y. Mansour, D. Wang, L. D. Zuck. Reliable Communication over Unreliable Channel. J. of the ACM 40(5):1087-1107, 1993.

No context found.

Afek, Attiya, Fekete, Fischer, Lynch, Mansour, Wang, and Zuck, Reliable Communication Over Unreliable Channels, JACM, vol. 41, no. 6, 1994, pages 1267--1297.

No context found.

Y. Afek, H. Attiya, A. D. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6):1267--1297, 1994.

No context found.

Y. Afek, H. Attiya, A. D. Fekete, M. Fischer, N. Lynch, Y. Mansour, D. Wang, and L. Zuck. Reliable communication over unreliable channels. Journal of the ACM, 41(6):1267--1297, 1994.

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