G. Neiger and S. Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11(3):374419, September 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....in [13] that one can transform many properties of one model into properties of another model. Another way to view such transformation is, that one transforms a distributed algorithm tolerant to certain failure types into an algorithm tolerant to some other, or, more failure types. Neiger and Toeug [14] describe transformations that convert a distributed algorithm tolerant of crash failures into an algorithm tolerant of omission or even arbitrary failures. The main differences to our approach are the following. First, we consider omission and performance failure, i.e. our failure model is ....

....algorithm tolerant of crash failures into an algorithm tolerant of omission or even arbitrary failures. The main differences to our approach are the following. First, we consider omission and performance failure, i.e. our failure model is weaker than the general omission failure model of [14] and stronger than their arbitrary failure model. The general omission transformation protocol of [14] is similar to the fault tolerant transformation protocol (Section 5.4.3) The message complexity of [14] are wo broadcasts ( Io unicasts) per round, where, is the number of processes. ....

[Article contains additional citation context not shown here]

G. Neiger and S. Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, (11):374--419, 1990.

....of computation that correspond to an increasingly stronger adversary. The idea of obtaining a protocol for an inhospitable environment by performing a series of transformations was perhaps rst employed in connection with fault tolerance. See, for example, the automatic translations discussed in [27]. Theoretical papers on cryptographic protocols often presume the existence of transformations from a more hostile model to a less hostile one. An author might, for instance, assume reliable links rather than our more realistic Fair Links, in order to ignore the complications of message ....

G. Neiger and S. Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11:374419, 1990.

....a set of processes propose a value and then they have to decide on a common value. A distributed consensus server can provide a basic building block to implement other fault tolerant services [8, 3, 13, 9] More formally, the consensus problem can be specified in terms of the following properties [12]: Termination: Every correct process eventually decides some value. Validity: If a process decides v, then v was proposed by some process. Agreement: No two processes decide differently. In this work we are interested in a solution for the consensus problem which can provide support for ....

G. Neiger and S. Toueg. "Automatically Increasing the Fault-Tolerance of Distributed Systems. Journal of Algorithms, 11(2):374--419, Sept. 1990.

....The complexity of protocol design is thus managed through a form of divide and conquer at each step, a designer focuses on a small number of additional attacks now permitted in the newly weakened system model. This idea of deriving a protocol is not new. For example, it has been proposed in [9, 26, 102, 81, 3, 4] as a way to obtain fault tolerant protocols for di#erent failure models. Prior work proposes mechanical transformations between protocols for di#erent system models. Although theoretically interesting, the mechanical transformations too often yielded unnecessarily complex and ine#cient protocols. ....

G. Neiger and S. Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11(3):374--419, September 1990.

.... of crash failures can be implemented on top of a system subject to more complex failures by running an appropriate software protocol [8] Second, techniques are available to automatically translate a protocol that tolerates crash failures into protocols that tolerate larger classes of failures [16]. 2.2 Fair Exchange Systems We view an exchange as a set of interactions among three processes: PA , PB and P TTP , which are running on behalf of player A, player B, and the TTP, respectively. Although the TTP usually serves multiple exchanges at the same time, we assume that different TTP ....

G. Neiger and S. Toueg. Automatically increasing the fault-tolerance of distributed systems. In Proceedings Seventh ACM Symposium on Principles of Distributed Computing, Toronto, Ontario, August 1988.

.... of crash failures can be implemented on top of a system subject to more complex failures by running an appropriate software protocol [8] Second, techniques are available to automatically translate a protocol that tolerates crash failures into protocols that tolerate larger classes of failures [15]. 2.3 Fairness loss owing to failures Process crashes can cause the following types of fairness loss in a fair exchange protocol modeled by Figure 3. Here we assume that no data item is stored to the stable storage during an exchange. Although an exchange protocol with an o line TTP is quite ....

G. Neiger and S. Toueg, Automatically increasing the fault-tolerance of distributed systems, in: Proceedings Seventh ACM Symposium on Principles of Distributed Computing, Toronto, Ontario, August 1988.

.... LSP82] Afterwards, the problem specifications that have been studied were often non uniform specifications, even in the setting of benign failures: for example, numerous results have been stated for consensus [FLP85, DDS87, DLS88, DRS90, DM90, CT96] and only a few are about uniform consensus [DS84, NT90, Lyn96]. Interestingly, Guerraoui [Gue95] showed that in most partially synchronous systems where processes may commit only crash failures, any algorithm that solves consensus also solves uniform consensus. In such systems, there is thereby no harm to solve consensus instead of uniform consensus. On the ....

....is, and so is harder than consensus. In the omission failure model, the comparison between the two problems is far less immediate. Perry and Toueg [PT86] exhibited consensus algorithms that tolerate any number of faulty processes. For uniform consensus, we can use the translation given in [NT90] which translates any algorithm tolerant of crash failures into one tolerant of omission failures. The translation works only if a minority of processes may fail. As long as this assumption holds, any algorithm that solves uniform consensus in the crash failure model is converted by means of this ....

[Article contains additional citation context not shown here]

G. Neiger and S. Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11(3):374--419, 1990.

....1 processes fail, nonfaulty participating processes output distinct values in the range f1; pg. 4.1. Algorithm There is a synchronous algorithm solving strong renaming [13] it requires log 2 p 2 rounds and tolerates p Gamma 1 crash failures. The transformation of Neiger and Toueg [15] can be applied to make the algorithm tolerate the same number of send omission failures, by doubling the number of rounds. The timeout technique of Attiya et al. 3, Sec. 4] is used to simulate a synchronous algorithm tolerating send omission failures by a semi synchronous algorithm tolerating ....

Neiger, G., and Toueg, S. Automatically increasing the fault-tolerance of distributed algorithms. J. Algorithms 11, 2 (1990), 374--419.

....the protocol is non blocking and has failover time f(2ffi ) for arbitrarily small and positive , so the lower bounds on blocking time and failover time are also tight. Most of the protocols for the various kinds of omission failures can be obtained by applying translation techniques [10] to the protocol for crash failures outlined above. These techniques re implement the message send and receive routines in such a way that a faulty server can detect its failure to send or receive a message and halt. All of the omission failure protocols obtained in this fashion have failover ....

Gil Neiger and Sam Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11(3):374--419, September 1990.

....servers and so Theorem 2 is tight. Furthermore, it is nonblocking and so incurs no additional delay. It has the failover time 2f5 r for arbitrarily small and positive r, and so Theorem 10 is tight. Most of our protocols for the various kinds of omission failures apply translation tech niques [17] to the protocol for crash failures outlined above. These techniques ensure that a faulty server detects its own failure and halts, thereby translating a more severe failure to a crash failure. The translations of [17] assume a round based protocol. Since our crash failure protocol is not ....

....for the various kinds of omission failures apply translation tech niques [17] to the protocol for crash failures outlined above. These techniques ensure that a faulty server detects its own failure and halts, thereby translating a more severe failure to a crash failure. The translations of [17] assume a round based protocol. Since our crash failure protocol is not round based, we must modify the translations so that a server can send and receive messages at any time rather than just at the beginning or the end of a round. All of these resulting omission failure protocols have failover ....

Gil Neiger and Sam Toueg. Automatically increasing the fault-tolerance of distributed systems. In Proceedings of the Seventh ACM Symposium on Principles of Distributed Computing, pages 248-262, Toronto, Ontario, August 1988. ACM SIGOPS-SIGACT.

....7 and 9 show that # satisfies # accuracy and # completeness, respectively. 10 Related Work The difference between the concepts of Agreement and Uniform Agreement was first pointed out in [Had86] in a comparison of Consensus versus Atomic Commitment. The term Uniform was introduced in [GT89, NT90], where it was studied in the context of Reliable Broadcast. In these papers, it is shown that with send and receive omission failures, URB can be solved if and only if a majority of processes are correct. BCBT96] consider systems with process crashes and fair (lossy) links, and addresses the ....

Gil Neiger and Sam Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11(3):374--419, 1990.

....in some applications, such as atomic commitment in distributed databases [BT93, Gra78, Had86] For such applications, a stronger version of reliable broadcast is more suitable, namely, uniform reliable broadcast which satisfies Uniform Integrity, Validity (Section 3. 2) and: Uniform Agreement [NT90] If any process delivers a message m, then all correct processes eventually deliver m. A quiescent implementation of uniform reliable broadcast can be obtained using quiescent implementations of reliable broadcast, and of quasi reliable send and receive between every pair of processes. ....

Gil Neiger and Sam Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11(3):374--419, 1990.

..... Uniform Validity: If a process decides v then some process previously proposed v. Agreement: Good processes do not decide different values. Termination: If all good processes propose a value, then they all eventually decide. A stronger version of consensus, called uniform consensus [9], requires: Uniform Agreement: Processes do not decide different values. The above specification allows a process to decide more than once. However, with Agreement, a good process cannot decide two different values. Similarly, with Uniform Agreement, no process (whether good or bad) can ....

Gil Neiger and Sam Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11(3):374--419, 1990.

....uses 3S. Since 3P 3S, this algorithm also solves Consensus using 3P. Our Consensus algorithms actually solve a stronger form of Consensus than the one specified in Section 5: They ensure that no two processes, whether correct or faulty, decide differently a property called Uniform Agreement [NT90] The Consensus algorithm that uses S tolerates any number of failures. In contrast, the one that uses 3S requires a majority of correct processes. We show that to solve Consensus this requirement is necessary even if one uses 3P, a class of failure detectors that is stronger than 3S. Thus, our ....

Gil Neiger and Sam Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11(3):374--419, September 1990. 44

....of the nonfaulty processors are required to choose the same action (such as deciding on an output bit) from a set of actions and to perform that action simultaneously. Instances of simultaneous choice problems include simultaneous versions of many well known problems, such as Reliable Broadcast [19], Byzantine Agreement [14,20] and Distributed Firing Squad [1,4,22] For example, in Byzantine Agreement, each processor starts with an input bit and chooses an output bit. All nonfaulty processors must choose the same output bit and this bit must be some processor s input bit. In Simultaneous ....

.... guarantees that all processors (whether faulty or nonfaulty) that reach a decision to commit or abort a specific transaction must reach the same decision [9,15] Other problems requiring that faulty processors behave consistently with nonfaulty processors appear in works by Neiger and Toueg [19] and by Gopal and Toueg [8] Given that consistency is sometimes possible in benign failure models, it is interesting to 1 Common Knowledge and Consistent Simultaneous Coordination ask whether consistency is always possible. In this paper, we define consistent simultaneous choice problems, in ....

[Article contains additional citation context not shown here]

Gil Neiger and Sam Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11(3):374--419, September 1990.

....different implementation of broadcast and deliver, and in all but one case optimal protocols are constructed. The protocols for crash and crash link failures show that all the corresponding lower bounds are tight. The protocol for general omission failures uses a translation technique similar to [8], and demonstrates that our lower bounds for general omission failures are tight, except for the bound on blocking time when f = 1. However, for this special case we have derived a different protocol (not described in this paper) having optimal blocking time . In all failure free runs of this ....

....the procedures for general omission failures is given in Figs. 8 and 9, except delivery process which is the same as Fig. 6. Whenever, we say that a site fifo delivered M , we mean that the procedure fifo deliver was called with M . These procedures were developed using a technique similar to [8] (although modified to work in our non round based model) which requires n s 2f and d = 2ffi. procedure initialize(k) statek : Rqueuek : Dqueuek : ffl 8i : Faultyk [s i ] false last sentk : 8j :expectedk [j] 0 procedure broadcast(M; k) time : current time fifo broadcast(init; ....

Gil Neiger and Sam Toueg. Automatically increasing the fault-tolerance of distributed systems. In Proceedings of the Seventh ACM Symposium on Principles of Distributed Computing, pages 248--262, Toronto, Ontario, August 1988. ACM SIGOPS-SIGACT.

....Coan s translations do not apply to synchronous systems. Other researchers have considered synchronous systems. Hadzilacos [14] developed a technique to translate a restricted class of algorithms tolerant of crash failures into ones that tolerate send omission failures. Neiger and Toueg [23] gave a translation from crash failures to send omission failures which translated a more general class of algorithms. While more general, their translation has higher round complexity than that of Hadzilacos. Srikanth and Toueg [28] showed how algorithms that use message authentication to ....

....While more general, their translation has higher round complexity than that of Hadzilacos. Srikanth and Toueg [28] showed how algorithms that use message authentication to mitigate arbitrary failures can be transformed into ones that do not require message authentication. Neiger and Toueg [23] developed a family of translations; some translate from crash to general omission failures, while others translate from general omission failures to arbitrary failures. They gave one translation from crash to general omission, and two translations from general omission to arbitrary failures. The ....

[Article contains additional citation context not shown here]

G. Neiger and S. Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11(3):374--419, September 1990. 83

No context found.

G. Neiger and S. Toueg. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms, 11(3):374419, September 1990.

No context found.

G. Neiger and S. Toueg. "Automatically Increasing the Fault-Tolerance of Distributed Systems". Journal of Algorithms, 11(2):374--419, 1990.

No context found.

G. Neiger and S. Toueg. "Automatically Increasing the FaultTolerance of Distributed Systems". Journal of Algorithms, 11(2):374--419, 1990.

No context found.

NEIGER, G. AND TOUEG, S. 1990. Automatically increasing the fault-tolerance of distributed algorithms. Journal of Algorithms 11, 3, 374--419.

No context found.

G. Neiger and S. Toueg. Automatically increasing the faulttolerance of distributed algorithms. Journal of Algorithms, 11(3):374--419, 1990.

No context found.

Neiger, G. and Toueg, S. (1990) Automatically increasing the fault-tolerance of distributed algorithms. J. Algorithms, 11, 374--419.

No context found.

G. Neiger and S. Toueg. Automatically increasing the faulttolerance of distributed algorithms. J. Algorithms, 11(3):374-- 419, Sept. 1990.

No context found.

Gil Neiger and Sam Toueg. Automatically increasing the fault-tolerance of distributed algorithms. J. Algorithms, 11(3), pages 374--419, September 1990.

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