M. Fischer, N. Lynch, and M. Paterson. Impossibility of distributed commit with one faulty process. J. of the ACM, 32(2), April 1985.  Home/Search   Document Not in Database   Summary   ACM   TOC   Related Articles   Check

....and an adversary scheduler, which tries to prevent them from doing so. At each step, the scheduler decides which processor operates next. How much information should be made available to the adversary for making this decision Most previous papers assume a dynamic adversary scheduler (e.g. [12, 10, 11, 1, 3, 14, 4]) having full knowledge of the entire state of the system at all previous steps. In the present work, we make a distinction between two separate components of the system s history: the control history and the content history. The control history includes all information regarding the dynamics of ....

....do not use division, where an interrupt may depend on the computed value. Similarly, the time to load and store data from memory does not depend on the data. Thus, the value oblivious adversary captures all the real world sources of asynchrony. Previous Related Work. Fischer, Lynch, and Paterson [12] prove the impossibility of deterministic asynchronous consensus in the message passing model, even if only one processor fails. Chor, Israeli, and Li [10] and Loui and AbuAmara [15] show that the same holds in the shared memory model (see also [11] Randomized solutions for the standard, full ....

M.J. Fischer, N.A. Lynch, and M.S. Paterson. Impossibility of distributed commit with one faulty process. Journal of ACM, 32(2):374--382, April 1985.

....to prevent consensus by controlling the processors speeds. Technically, at each instant in time, the adversary determines the next processor to operate. In order to make scheduling decisions, the adversary is typically assumed to know the entire state of the system at all previous times (see [15, 13, 14, 1, 3, 16, 4]) Fischer, Lynch, and Paterson [15] show that any consensus algorithm must be randomized. They prove the impossibility of deterministic asynchronous consensus, even if only one processor may fail. A subsequent sequence of papers reduces the complexity of randomized consensus from exponential ....

....Technically, at each instant in time, the adversary determines the next processor to operate. In order to make scheduling decisions, the adversary is typically assumed to know the entire state of the system at all previous times (see [15, 13, 14, 1, 3, 16, 4] Fischer, Lynch, and Paterson [15] show that any consensus algorithm must be randomized. They prove the impossibility of deterministic asynchronous consensus, even if only one processor may fail. A subsequent sequence of papers reduces the complexity of randomized consensus from exponential [1] to polynomial [3] to O(n log n) ....

[Article contains additional citation context not shown here]

M. J. Fischer, N. A. Lynch, and M. S. Paterson. Impossibility of distributed commit with one faulty process. Journal of ACM, 32(2):374--382, April 1985.

....of failure tolerated by the algorithm. If the processes can exhibit stopping failures then the Termination requirement is too strong and must be replaced by Wait free termination: All initialized and non failed processes eventually decide. Unfortunately, the fundamental impossibility result of [7] demonstrates that there is no deterministic algorithm for achieving waitfree agreement in the asynchronous distributed model with communication via shared read write variables even in the presence of one stopping failure 2 . One solution is to use randomization, which necessitates a weaker ....

M. Fischer, N. Lynch, and M.Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(5):374-382, 1985.

....wait free consensus protocol, one in which the behavior of the processors is predictable in advance. In fact, in a wide variety of asynchronous models it has been shown there is no deterministic consensus protocol that works against a scheduler that can stop even a single processor [CIL87, DDS87, FLP85, Her91, LAA87, TM89] Though this result is usually proved using more general techniques, when only single writer registers are used it has a simple proof that illustrates many of the problems that arise when trying to solve wait free consensus. Imagine that two processors A and B are trying to ....

....taxonomy can be found in [LL90] Dolev et al. DDS87] classify a large collection of message passing models and show which are capable of solving consensus deterministically. Among these many models, one has traditionally been associated with solving asynchronous consensus [BND89, BT83, CM89, FLP85] In this model, the adversary is allowed to (i) stop up to t processors and (ii) delay messages arbitrarily. Unfortunately, a simple partition argument shows that in this model one cannot solve consensus even with a randomized algorithm if at least n=2 processors can fail [BT83] Intuitively, ....

Michael J. Fischer, Nancy Ann Lynch, and Michael S. Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(2):374--382, April 1985.

....and thus it cannot be outwitted by encryption schemes. The adversary cannot, however, predict future coin flips. Against such a powerful adversary, it may seem surprising that consensus can be achieved by a simple protocol in polynomial expected time. 2 Related Work Fischer, Lynch, and Paterson [20] show that there is no consensus protocol for two processes that communicate by asynchronous messages. Dolev, Dwork, and Stockmeyer [15] and Dwork, Lynch, and Stockmeyer [16] give a comprehensive analysis of the circumstances under which consensus can be achieved by message passing. Randomized ....

M. Fischer, N.A. Lynch, and M.S. Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(2), April 1985.

....are chosen to simulate the conditions under which the protocol is required to operate. It has been known for some time that the ability of the scheduler to stop even a single processor is sufficient to prevent consensus from being solved by a deterministic algorithm using only atomic registers[17, 14, 11, 9, 15]. Chor, Israeli, and Li[9] 2 showed that it is possible to solve consensus using globally visible atomic coin flips as primitive actions. Since then a number of protocols have been put forward[1, 6, 7] which achieve consensus with only local coin flip operations. These protocols work even against ....

M. Fischer, N.A. Lynch, and M.S. Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(2), April 1985.

....failure of its transaction This is what is called the consensus problem: we want to make two processors agree on a common value. This question is a particular case of a more general one which is to know what this concurrent machine can compute. In the case of two processors, this was solved in [22]: it is by no means possible to solve the consensus problem on our simple machine. But we had to wait until quite recently for a characterization of what can be computed on our asynchronous machine with n processors (for any n) This has been done by methods borrowed from combinatorial algebraic ....

.... derive from this work (see for instance [41] We refer the reader to other articles in this area, in particular, the book on distributed algorithms [46] other articles by Herlihy et al. 37] 38] 39] some slightly different methods, but still geometric in nature, in [2] 5] 6] 11] [22] (which originated this field of research, starting with graph theoretical arguments) 51] Also some ideas about classifying data structures according to what protocols they manage to solve are described in [42] 43] 53] This should be related to problem (CS2) Some links with directed ....

M. Fisher, N. A. Lynch, and M. S. Paterson, Impossibility of distributed commit with one faulty process, Journal of the ACM 32 (1985), no. 2, 374--382.

....of failure tolerated by the algorithm. If the processes can exhibit stopping failures then the Termination requirement is too strong and must be replaced by: Wait free termination: All initialized and non failed processes eventually decide. Unfortunately, the fundamental impossibility result of [7] demonstrates that there is no deterministic algorithm for achieving wait free agreement in the asynchronous distributed model with communication via shared read write variables even in the presence of at most one stopping failure 2 . One solution is to use randomization, which necessitates a ....

M. Fischer, N. Lynch, and M.Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(5):374-382, 1985.

....data types, including stacks, priority queues, fetch add registers [16] and others. This argument also yields lower bounds for decision problems that solve strong renaming, including assignment [4] and order preserving renaming [2] Elsewhere [14] a similar kind of reduction (to consensus [6, 10]) was used to derive impossibility results for wait free concurrent objects in the asynchronous shared memory model. Our results here show that reduction to a decision problem can be adapted to yield complexity as well as impossibility results. How tight are these lower bounds for concurrent ....

M. Fischer, N.A. Lynch, and M.S. Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(2), April 1985.

....in the sense that any RMW can be atomically implemented from a 3 value atomic RMW in a wait free fashion using bounded memory. The Sticky Bit can be viewed as a version of consensus, which has proven to be a valuable tool in understanding the limitations of asynchronous distributed systems [5, 6]. On the other hand, the definition of the Sticky Bit is memory oriented, which makes it a convenient alternative to consensus in the context of shared memory systems. A Sticky Bit object can be easily constructed from two safe bits and a single initializable object that implements a wait free ....

....In addition, it supports a non atomic Flush operation. ffl Flush Sets the value to . This operation is non atomic in the sense that any other operation that overlaps it produces unpredictable results. A consensus protocol was defined in the seminal paper of Fisher, Lynch, and Paterson [6] to be a protocol where each processor has a 1 bit input and produces a 1 bit output which confirms to two conditions. First, all produced outputs are the same, and second, if the output is v then there is at least one participating processor whose input is v. The consensus protocol can be ....

M. J. Fischer, N.A. Lynch, and M.S. Paterson. Impossibility of Distributed Commit with One Faulty Process. J. ACM, 32(2), April 1985.

....providing a characterization of solvability of decision problems in the style of [8] which, for some of the models, is given for the first time. 1 Introduction For almost two decades now, the consensus problem has played a central role in the study of fault tolerant distributed computing, e.g. [23, 13, 12, 10, 14, 20, 16, 8, 9]. It has clearly received the greatest amount of attention in the theoretical literature on distributed computing, and has been studied in a large variety of models and under many types of failure assumptions. Work on different variants often in This work has been supported by a Helen and Milton ....

....infinite run. Obviously, in some models it is possible to determine that a process is faulty by considering only a prefix of the run, sometimes even a single state. In other cases, however, it is impossible to determine this after a finite amount of time. See, e.g. the asynchronous systems in [14]. 3 Consensus and Connectivity In the classical binary consensus problem [23] processes start with binary values and have to decide on binary values satisfying three conditions. Decision, requires that every nonfaulty process eventually decides; Agreement requires that the decisions are ....

[Article contains additional citation context not shown here]

M. Fischer, N.A. Lynch, and M.S. Paterson, "Impossibility of distributed commit with one faulty process," Journal of the ACM, 32(2), April 1985.

....Parte 1. Modelo de memoria compartida. 2. Problemas de consenso solucionables (consenso seguro [5, 16] y no solucionables. 3. Problema de los conjuntos ordenados y el modelo de la foto at omica inmediata [5, 6] 4. Imposibilidad de consenso y estructura topol ogica del c omputo libre de espera [3, 5, 9]. 6 5 Tareas y Evaluaci on Hubo cinco tareas: dos referentes a la primera parte (una de ellas dedicada al problema del arbol generador de peso m inimo, y la otra al resto) y una a cada una de las otras partes. Incluyeron ejercicios sencillos para ayudar a asimilar el material visto en clase, ....

M. Fischer, N.A. Lynch, and M.S. Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(2), April 1985.

....that the simplicial complex associated with any wait free protocol using read write memory has a remarkable topological property: it has no holes in any dimension. We exploit this simple property to derive our impossibility results. In a fundamental paper in 1985, Fischer, Lynch, and Paterson [19] showed that there exists a simple task that cannot be solved in a message1 passing system if even one process may fail by halting (or may be infinitely slow) This result showed that the notion of asynchronous computability differs in important ways from conventional notions of computability ....

....each of the N = n 1 processes starts with a private input value, communicates by reading and writing variables in shared memory, and halts with a private output value. For example, in the well known binary consensus task, the processes have binary inputs, and must agree on some process s input [19]. A protocol is a program that solves a decision task. A protocol is wait free if it guarantees that every non faulty process will finish in a bounded number of steps, no matter how many processes fail. In this paper, we investigate protocols in which processes communicate by reading and writing ....

[Article contains additional citation context not shown here]

M. Fischer, N.A. Lynch, and M.S. Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(2), April 1985.

....between any participants and T is over a private channel. We will first concentrate on the case of two participants, A and B. The network model we consider is the same as in [3] Namely, an asynchronous communication model with no global clocks, where messages can be delayed arbitrarily [27], but with messages sent between correct participants and the trusted third party guaranteed to be delivered eventually. In general, we assume 2 Depending on the proof structure, this probability might be a constant fraction, in which case the standard method of repeating the proof can be ....

....either universally agreed upon, or is part of the contract m itself. For instance, valid evidence may include either valid signatures by A and B on m, or valid signatures by A and B on the concatenation of m and the public key of T , signed together by T . Obviously, in the asynchronous model of [27], an adversary may prevent a contract from being signed simply by delaying all messages between players. Thus in order to force contract signing protocols to be non trivial, we specify a completeness condition using a slightly restricted adversary. An optimistic contract signing protocol is ....

M. Fischer, N. Lynch, and M. Paterson. Impossibility of distributed commit with one faulty process. J. ACM, 32(2), 1985.

No context found.

M. Fischer, N.A. Lynch, and M.S. Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(2), April 1985.

No context found.

M. Fischer, N. Lynch, and M. Paterson. Impossibility of distributed commit with one faulty process. J. of the ACM, 32(2), April 1985.

No context found.

M. Fischer, N. Lynch, and M. Paterson. Impossibility of distributed commit with one faulty process. J. of the ACM, 32(2), April 1985.

No context found.

M. Fischer, N. Lynch, and M. Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(5):374--382, 1985. 9

No context found.

M. Fischer, N. Lynch, and M. Paterson. Impossibility of distributed commit with one faulty process. J. of the ACM, 32(2), April 1985.

No context found.

M. J. Fischer, N. A. Lynch, and M. S. Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(2):374-382, April 1985.

No context found.

M. Fischer, N. Lynch, and M. Paterson. Impossibility of distributed commit with one faulty process. J. of the ACM, 32(2), April 1985.

No context found.

M. Fischer, N. Lynch, and M. Paterson. Impossibility of distributed commit with one faulty process. J. of the ACM, 32(2), April 1985.

No context found.

M. Fischer, N. Lynch, and M. Paterson. Impossibility of distributed commit with one faulty process. Journal of the ACM, 32(5):374-382, 1985.

No context found.

M. Fischer, N. Lynch, et al. Impossibility of distributed commit with one faulty process. J. of the ACM, 32(2), Apr. 1985.

No context found.

M. Fischer, N.A. Lynch, and M.S. Paterson. Impossibility of Distributed Commit with one faulty process. Journal of the ACM,32(2), April 1985.

First 50 documents  Next 50

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