Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems, ACM Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....that have been issued to modify the service state. So, there are two problems to solve. 1) Disseminate the updates to the nodes that have a copy of the service state. And (2) apply the updates in the same order to each copy. The first problem can be solved by using a reliable multicast primitive [18]. The second problem is more difficult to solve. The nodes have to agree on a common value, namely, the order in which they will apply the updates. This well known problem (namely, the Atomic Broadcast problem) is actually a classical Agreement problem. It appears that any agreement problem can ....

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems (Second Edition), acm Press (S. Mullender Ed.), New-York, 1993, pp. 97-145.

....which solves the problem of the consensus using the OS class. Rounds Number I CCP 3 Messages number 3(n1) Figure 6: The complexity of the consensus solution(using 05) Note that we must add the number of the exchanged messages, required by an execution of the reliable broadcast primitive [7], to the total number of the exchanged messages. 6 Conclusion Several atomic commitment protocols have been proposed in the literature, unfortunately most of them are blocking in case some participants crash. In this paper we gave a comparative study between the two phase commit protocol and a ....

Hadzilacos V., Toueg S. Reliable Broadcast and Related Problems. Distributed Systems (Second Edition), ACM Press (S. Mullender 1993, p. 97-145. Ed.), New-York,

....provides it with the decided value (at line 12 or 15) It is possible that distinct processes do not decide during the same round. To prevent a process from blocking forever (i.e. waiting for a value from a process that has already decided) a process that decides, uses a reliable broadcast [4] to disseminate its decision value. To this end, the Consensus function is made of two tasks, namely, T1 and T 2. T1 implements the previous discussion. Line 12 and T2 implement the reliable broadcast. 3.2 Proof 3.2.1 Validity Theorem 1 If a process p i decides v, then v was proposed by some ....

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems, acm Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.

....synchronous system does not allow for more efficient solutions that an asynchronous system equipped with a perfect failure detector. A solution to the GDC problem can be designed in executing n parallel instances (one for each process) of the Uniform Terminating Reliable Broadcast (U TRB) problem [5]. For each U TRB instance (an instance is defined with respect to a specific sender process) a value v is delivered to each process. If v = a predetermined default value) then the sender has necessarily crashed. If v 6= then v is the initial value of the sender process. Moreover, if a ....

....is delivered to each process. If v = a predetermined default value) then the sender has necessarily crashed. If v 6= then v is the initial value of the sender process. Moreover, if a process gets v, then no process gets a different value. A U TRB protocol requires (t 1) computation rounds [5]. The use of U TRB as a subroutine allows designing a GDC protocol that inherits its number of rounds. A protocol described in [10] achieves early stopping by solving the Terminating Reliable Broadcast problem in min(f 2; t 1) rounds, in a synchronous distributed system where processes can ....

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems, acm Press (S. Mullender Ed.), New-York, pp. 97-145, 1993. 18

....that states the order (if any) in which messages have to be delivered. Usually the ordering property depends on the message sending order. fifo, causal A previous version of this paper has appeared in [1] 2 order (co) 3,4] and total order (to) 3,5] are the most encountered ordering properties [6]. Another type of communication service is offered by CSP like languages (e.g. the Occam distributed programming language) This communication type assumes reliable processes and provides the so called rendezvous (rdv) communication paradigm [7 9] also called synchronous communication. This ....

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems, acm Press (S. Mullender Ed.), New-York, pp. 97-145, (1993). 13

....its destination processes despite possible failures. An ordering property states the order in which messages have to be delivered; usually this order depends on the message sending order. fifo, causal order (co) 4, 18] and total order (to) 4, 7, 11] are the most encountered ordering properties [12]. Reliability and ordering properties can be combined to give rise to powerful communication primitives such as Atomic Broadcast [4] or Atomic Multicast to asynchronous groups [9] The ISIS system [4] pioneered research in this domain. Other systems that have proposed powerful communication ....

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems, acm Press (S. Mullender Ed.), New-York, pp. 97-145, (1993).

....and Total Order. ffl Validity: If a process A delivers a message m, then some process has Abroadcast m. ffl Integrity: Let p be the delivery sequence at a given process p. A message m appears at most once in p . 1 We actually consider the definition of the Uniform Atomic Broadcast problem [11]. 4 ffl Termination: For any message m, 1) if the process that issues A broadcast(m) eventually remains permanently up, or (2) if a process A delivers a message m, then all processes that eventually remain up A deliver m. ffl Total Order: Let p be the sequence of messages A delivered to ....

V. Hadzilacos and S. Toueg, Reliable Broadcast and Related Problems. Chapter 4, Distributed Systems (2nd Edition), acm Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.

....is binary when the set V consists of only two values [2, 18] It is multivalued when the set V can be arbitrarily large. 3 A Randomized Multivalued Consensus Protocol 3. 1 Preliminary: Reliable Broadcast The proposed randomized protocol uses the Reliable Broadcast communication primitives [12], namely, R Broadcast(m) and R Deliver(m) When a process issues R Broadcast(m) we say that it R broadcasts m. Similarly, when a process issues R Deliver(m) we say that it R delivers m. Reliable Broadcast is defined by the following set of properties [12] 2 ffl Termination: If a correct ....

....Broadcast communication primitives [12] namely, R Broadcast(m) and R Deliver(m) When a process issues R Broadcast(m) we say that it R broadcasts m. Similarly, when a process issues R Deliver(m) we say that it R delivers m. Reliable Broadcast is defined by the following set of properties [12]: 2 ffl Termination: If a correct process R broadcasts m, then any correct process R delivers m (no message from a correct process is lost) ffl Uniform Agreement: If a process R delivers m, then any correct process R delivers m (no message R delivered by a correct or not process is missed by ....

[Article contains additional citation context not shown here]

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems, acm Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.

.... p k is correct, it should verify ; if is found to be a valid input, p k should then construct message m k ( that contains and broadcast m k ( to other servers for identical ordering (we call this p k forwarding to its peers ) This ordering is typically achieved by Atomic Broadcast protocols [2, 6, 13]. Say, a correct p i makes an ordered delivery of m k ( it should not process the contained input straightaway unless is locally verified, since the forwarding process may be faulty: p k may not have verified properly or, even worse, may be replaying an old input as a newly received input or ....

....of the state of the service they implement. This can be obtained by requiring correct processes to order the client inputs in the same way before processing them according to this order [13] This well known problem is the Atomic Broadcast problem defined by the following set of properties 2 [6]: ffl AB Validity. An input issued by a correct client is ordered by each correct process. ffl AB Integrity. Any input ordered by a correct process comes from a client. Moreover, any input is ordered at most once by a correct process. ffl AB Agreement. If an input is ordered by a correct ....

[Article contains additional citation context not shown here]

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems, ACM Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.

....processes voted yes) Finally, the obligation property eliminates the trivial solution where the decision value would be abort even when the situation is good enough to commit. 3. 3 Ordered Reliable Broadcast Here we are concerned with a communication primitive which is an agreement problem [14]. It is question of broadcast, more precisely of Ordered Reliable Broadcast (ORB) 10 . We define this communication primitive in an incremental way. We first introduce a Broadcast primitive, then a Reliable Broadcast primitive, and finally the Ordered Reliable Broadcast primitive. In all cases, ....

Hadzilacos V., Toueg S. Reliable Broadcast and Related Problems. In Distributed Systems (Second Edition), acm Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.

....been generalized by Guerraoui and Schiper in [17] where they introduce a methodolgy to define very general Consensus services in the context of unreliable asynchronous distributed systems. Actually, this paper can be seen as a state of the art introduction (based on previous works described in [1, 4, 14, 16, 17, 19, 23]) to a class of agreement problems and fault tolerance techniques for distributed systems, with a special emphasis on the nbac problem. 2 Distributed Systems and Transactions 2.1 Distributed Systems A distributed system is composed of a finite set of sites interconnected through a communication ....

....primitives constitutes a fundamental point in the design of distributed systems. Among these primitives multicast is particularly important as it allows a process to disseminate a message m to a set P = fp 1 ; p 2 ; p n g of processes. We consider here three multicast primitives (see [1, 19] for in depth study of such communication primitives) Unreliable multicast: Multisend (m,P) Actually this primitive is a syntactical notation that abstracts: for each p 2 P do send(m) to p end do where send is the usual point to point communication primitive. Multisend (m,P) is the simplest ....

[Article contains additional citation context not shown here]

V. Hadzilacos and S. Toueg. Reliable Broadcast and Related Problems. In Distributed Systems (Second Edition), acm Press (S. Mullender Ed.), New-York, 1993, pp. 97-145.

.... multisend are unreliable: the channel can lose messages but it is assumed to be fair, i.e. if a message is sent infinitely often by a process p then it is received infinitely often by its receiver [13] When a message 1 We actually consider the definition of the Uniform Atomic Broadcast problem [10]. 2 The primitive multisend is actually a macro that allows a process p to send (by using the basic send primitive) a message to all processes (including itself) 5 arrives at a process it is deposited in its input buffer that is a part of its volatile memory. The process will consume it by ....

V. Hadzilacos and S. Toueg, Reliable Broadcast and Related Problems. Chapter 4, Distributed Systems (2nd Edition), acm Press (S. Mullender Ed.), New-York, pp. 97-145, 1993. 19

....no duplicates. Integrity: Let p be the delivery sequence at a given process p. A message m appears at most once in p . The termination property specifies the situations where a message m has to be A delivered. 1 We actually consider the definition of the Uniform Atomic Broadcast problem [10]. 4 . Termination: For any message m, 1) if the process that issues A broadcast(m) eventually remains permanently up, or (2) if a process A delivers a message m, then all processes that eventually remain up A deliver m. The total order property specifies that there is a single total order ....

V. Hadzilacos and S. Toueg, Reliable Broadcast and Related Problems. Chapter 4, Distributed Systems (2nd Edition), acm Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.

....and applying writes to the database are completely handled by the Bayou database managers. Applications simply issue read and write operations and observe the effects of eventual consistency. Applications can optionally request additional session guarantees that affect the observed consistency [18]. 2.3 Replica creation and retirement Bayou permits the number and location of replicas for a database to vary over time. While the replica placement policies are under the control of applications as discussed below in section 3.1, the mechanism for creating new replicas and retiring old ones is ....

....Proceedings 16th ACM Symposium on Operating Systems Principles, Saint Malo, France, October 1997, pages 288 301. 17] P. Reiher, J. Heidemann, D. Ratner, G. Skinner, and G. Popek. Resolving file conflicts in the Ficus file system. Proceedings Summer USENIX Conference, June 1994, pages 183 195. [18] D. B. Terry, A. J. Demers, K. Petersen, M. J. Spreitzer, M. M. Theimer and B. B. Welch. Session guarantees for weakly consistent replicated data. Proceedings Third International Conference on Parallel and Distributed Information Systems, Austin, Texas, September 1994, pages 140 149. 19] D. B. ....

[Article contains additional citation context not shown here]

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems (Second Edition), ACM Press (S. Mullender Ed.), New-York, 1993, pp. 97-145.

....Each process p i has an initial value v i that it proposes to the others, and all sound processes have to decide on a single value that has to be one of the proposed values. More precisely, the Consensus problem is de ned by the following three properties (we actually consider the Uniform version [10, 13] of the Consensus problem) Termination: Every sound process eventually decides some value. Uniform Validity: If a process decides v, then v was proposed by some process. Uniform Agreement: no two processes (sound or awed) decide di erently. 3.5 Enriching the Model to Solve Consensus ....

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. Chapter 4, Distributed Systems (2nd Edition), acm Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.

No context found.

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems, ACM Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.

No context found.

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems, acm Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.

No context found.

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems, acm Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.

No context found.

Hadzilacos V. and Toueg S., Reliable Broadcast and Related Problems. In Distributed Systems, ACM Press (S. Mullender Ed.), New-York, pp. 97-145, 1993.

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