K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Trans. on Computer Systems, 9(3):272-314, Aug. 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....exchange many additional messages and thus are impractical to use in high performance communication systems. Furthermore, these protocols are partially correct in the sense that they terminate with a probability that increases asymptotically to one as the protocol advances. Asymmetric protocols ([7, 9, 21]) use a centralized coordinator for ordering the messages. The problems with this method are in the serial bottleneck they create at the coordinating site, and with the costly handling of faults in case the coordinator crashes. Existing symmetric protocols ( 28, 6, 17, 24] require all machines ....

....ordering of messages have been designed that circumvent the impossibility result. Synchronous protocols ( 16, 12, 18] circumvent the impossibility result by explicitly assuming synchrony. Probabilistic protocols ( 4, 29, 8, 10, 13, 26] introduce random steps to the protocol. In other protocols ([9, 7, 6, 24]) the system halts and reconfigures when processors fail or join. In Transis, the impossibility result is circumvented in the automatic maintenance of dynamic membership ( 1] The impossibility result stems from the inability to distinguish between slow and faulty machines in the asynchronous ....

[Article contains additional citation context not shown here]

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Trans. Comput. Syst., 9(3):272--314, 1991.

....the timewheel group membership protocol. These two protocols along with a clock synchronization protocol [24] comprise the timewheel group communication system. The timewheel group communication system provides four unique characteristics that distinguish it from other group communication services [12, 8, 44, 52, 9, 3, 39, 4, 21, 50, 45, 19, 5, 7]. First, this system has been designed for a timed asynchronous distributed system model [18] Timed asynchronous distributed system model has been proposed recently. It allows the construction of dependable protocols that specify what outputs and state transitions should occur in response to ....

....an update. In particular, a user may broadcast one update with one pair of atomicity and order semantics, and the next update with another pair. Group communication systems proposed earlier have provided similar semantics. These include virtual synchrony [8] causal and atomic broadcasts [9, 39, 3], extended virtual synchrony [42] and uniform broadcast [47] An important contribution of the timewheel group communication system is that it has classified group communication semantics into logical classes. Also, although some of the group communication systems proposed earlier have provided ....

[Article contains additional citation context not shown here]

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Transactions on Computer Systems, 9(3):272--314, Aug 1991.

.... two schemes has emerged [9] active and passive (or primarybackup) replication [4] The active replication paradigm does not impose and exchange of messages among replicas, however it requires each replica to be deterministic and relies on the availability of a total order multicast primitive [3] from the communication subsystem. This allows a total order of requests incoming from distinct clients at each replica. To implement this primitive it is necessary a sort of coordination (i.e. synchronization) among the communication subsystem entities local at each object. Passive replication ....

.... time where a leader is one object that receives a support message, in a timely way, from a majority of objects [7] In the context of the asynchronous model, the many to many synchronization cen be realized by a service of group membership which provides each object with the same sequence of views [3, 16]. A view contains all non crashed objects. Therefore the primary can be selected by any object in the view using the same deterministic rule. Another method has been proposed by Guerraoui and Frolund in [8] They use the notion of Write Once Register [10] which is actually a simple extension of a ....

[Article contains additional citation context not shown here]

K. Birman, A. Schiper, and P. Stephenson, Lightweight Causal and Atomic Group Multicast, ACM Transactions on Computer Systems 9 (1991), no. 3, 272--314.

....In active replication [1] a client sends a request to a set of deterministic server replicas. Each replica 20 independently executes the request and sends back a reply to the client. To get linearizability, clients and servers must interact through a total order (or atomic) multicast primitive [17]. This primitive ensures that all server replicas process requests in the same order before failing. In passive replication [2] a particular primary replica serves all client requests. Upon receiving a request req, the primary processes the request, produces a result res and reaches a new state. ....

Birman K, Schiper A, Stephenson P. Lightweight causal and atomic group multicast. ACM Transactions on Computer Systems 1991; 9(3):272--314.

....techniques for reducing synchronization across replicas are used to scale a zero downtime EDE to support the large number of subscribers it must service. 1 Introduction Increasing a server system s performance, reliability, and fault tolerance by means of replication is common practice [2, 5, 6, 7, 9, 10, 22]. Performance improvements are attained by use of parallelism and concurrency [19] By using additional techniques for fault detection, masking and recovery, replication can deal with hardware failures, and with software failures caused by non determinism or by certain behavior at isolated ....

....at more constant levels. Hence, the second tier contributes towards our goal of perceived zero downtime and masked failures. 5 Related Work Using replication as a technique to increase system reliability and fault tolerance has been widely accepted among both researchers and in industry (e.g. [2, 5, 6, 7, 9, 10, 22]) As with our solution, most approaches rely on (1) a checkpointing mechanism through which they track each other s progress, and (2) a log of events that must be either undone or redone when a failure occurs [3, 12] The exchange of heartbeat traffic, either explicit or embedded in other ....

[Article contains additional citation context not shown here]

K. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Transactions on Computer Systems, 9(3), Aug. 1991.

....the next sections. In active replication, a client sends a request to a set of deterministic server replicas. Each replica executes independenfiy the request and sends back the reply to the client. To get linearizability, clients and servers must interact through a total order multicast primitive [6]. This primitive ensures that all the server replicas process requests in the same order before failing, i.e. protocols implementing total order multicast let replicas agree on the order of message deliveries. In passive replication, a particular primary replica serves all client requests. Upon ....

K. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Transactions on Computer Systems, 9(3):272 314, August 1991.

....through reliable channels. Communication is asynchronous and by exchanging messages. A failed process can later recover with its permanent storage intact and re join the system. Failures are detected using a (possibly unreliable) failure detector [CT96] A virtual synchronous multicast service [BSS91, Bir96, SR93] is used. This service delivers multicast messages and views. Views indicate which processes are perceived as up and connected. We assume a virtual synchrony with the following properties: 1) Strong virtual synchrony [FvR95] or sending view delivery [VKCD99] that ensures that messages are ....

K. P. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Transactions on Computer Systems, 9(3):272-314, August 1991.

....q (m 2 ) send q (m 3 ; r) ffl is a transitive relation. Therefore, r must deliver m 3 only after m 1 has been received and delivered. There are two fundamental approaches to implementing causal message delivery. The first is to add to each message m additional information [RST91,SS92, BSS91] that m s destination process uses to determine when m can be delivered. 46 message is piggybacked on message q r = m Figure 5.2: Implementing causal delivery order through piggybacking Using this approach, process r in Figure 5.1 would realize, when it receives m 3 , that it ....

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Transactions on Computer System, 9(3):272--314, August 1991.

....to exploit topology information to improve 3 the design of group communication protocols. Several other examples can be found in the literature. However, with a few notable exceptions, most examples were focused on the solution of a particular problem, and have put their emphasis on algorithmics[4, 7, 8, 16, 23, 32]. To our knowledge, NAVTECH was pioneer in the coherent and systematic use of topology awareness, through the definition of a generic architectural construct, the WANof LANs model. This enabled the application of topology awareness in a vertical manner, to practically all protocols developed for ....

....member of a group in a site, only one message is sent there, and then copied to all recipients. Likewise, when a site fails, a single run of a site failure detection algorithm needs to be executed, instead of having many runs executing in parallel as in some other earlier generation groups systems[33, 26, 7]. On the other hand, site group members process the send requests from their local senders, and run the protocols that ensure delivery to the members, with the requested guarantees. Senders are not full right members of a group. They get from the system just the necessary information and support ....

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM, Transactions on Computer Systems, 9(3), August 1991.

....this classification, message ordering can be built by: 1.senders(e.g. 1, 12, 11, 21, 16] in this case senders agree on the order to submit requests to the replicas; 2.destinations (e.g. 5, 10] in this case replicas agree on the order to process client requests; 3. external processes (e.g. [6, 8, 16, 20]) ordering may involve processes that are neither senders nor destinations. In algorithms of the first class clients must synchronize among them to get a total order. This violates Client Asynchrony. Algorithms of the second and third classes can be adapted to satisfy Client Asynchrony. For ....

....it is easy to see that it is not satisfied by algorithms falling in the second class. Algorithms of the third class are commonly designed assuming that the external process is elected among clients or servers that have to run in a partially synchronous distributed system. As example, in [6] processes elect a sequencer process that is in charge of defining total order. Election is based on a group membership service that is impossible to implement in asynchronous distributed systems [9] 3 Three tier (3T) active replication 3T active replication introduces a middle tier (midtier) ....

K. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Transactions on Computer Systems, 9(3):272-- 314, August 1991.

....the next sections. In active replication, a client sends a request to a set of deterministic server replicas. Each replica executes independently the request and sends back the reply to the client. To get linearizability, clients and servers must interact through a total order multicast primitive [6]. This primitive ensures that all the server replicas process requests in the same order before failing, i.e. protocols implementing total order multicast let replicas agree on the order of message deliveries. In passive replication, a particular primary replica serves all client requests. Upon ....

K. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Transactions on Computer Systems, 9(3):272--314, August 1991.

....application is the database state machine [21] which allows high performance replication of transactional databases. Implementation of total order multicast is however more costly than other forms of multicast due to the unavoidable additional latency. For instance, in a sequencer based protocol [5, 16] all processes (except the sequencer itself) have to wait for the message to reach the sequencer and for the sequence number to travel back before the message can be delivered. On the other hand, protocols based on causal history [18, 23, 10] can provide latency proportional to the interarrival ....

....terarrival time and low latency is desired, this requires the introduction of additional control messages. This is especially unfortunate in large groups and in wide area networks with limited bandwidth links. In some protocols, such as those based on consensus [7, 4] or on a sequencer [5, 16], the total order decided is the spontaneous ordering of messages as observed by some process. In addition, in local area networks (LANs) it can be observed that the spontaneous order of messages is often the same in all processes. The latency of total order protocols can therefore be masked (not ....

[Article contains additional citation context not shown here]

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Trans. Comput. Syst., 9(3), Aug. 1991.

....between them degrades Sirac is a joint laboratory of Institut National Polytechnique de Grenoble, INRIA and Universit Joseph Fourier. the performance of classical clock synchronization algorithms. A common or dering mechanism uses logical time [3] to order events according to the causal order [4]. The causal precedence relation induces a partial order on the events of a distributed computation. It is a powerful concept, which helps to solve a variety of problems in distributed systems like algorithms design, concurrency measurement, tracking of dependent events and observation [5] In ....

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. In ACM Transactions on Computer Systems, volume 9, pages 272-314, August 1991.

....adapted for mobile platforms thanks to features of message queuing (which ensures a certain reliability) or message ordering (which provides a way to reduce the non determinism) for instance. A common ordering mechanism uses logical time [2] to order events according to the causal order [3]. The causal precedence relation induces a partial order on the events of a distributed com putation. It is a powerful concept, which helps to solve a variety of problems in distributed systems like algorithms design, concurrency measurement, tracking of dependent events and observation. The ....

K. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. In ACM Transactions on Computer Systems, volume 9, pages 272-314, August 1991.

No context found.

Kenneth P. Birman, Andre Schiper and Pat Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Transactions on Computer Systems, Vol. 9, No 3, August, 1991, 272314.

No context found.

BIRMAN, K. P., SCHIPER, A., AND STEPHENSON, P. 1991. Lightweight causal and atomic group multicast. ACM Trans. Comput. Syst. 9, 3 (Aug.), 272--314.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Trans. on Computer Systems, 9(3):272--314, August 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Transactions on Computer Systems, 9(3):272--314, August 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Trans. on Computer Systems, 9(3):272--314, August 1991.

....The group communication protocol suite implemented in the Appia framework [26] is also strongly inspired by Ensemble. The membership service presented in [23] uses a token based approach as in Totem or RMP. 2.1 Monolithic Systems Isis. Isis was the first system to propose group communication [7, 8]. It is a monolithic primary partition system, i.e. when a network partition occurs, the computation can only proceed in one partition of the network, called the primary partition. The Isis architecture is depicted in Figure 1. The main layers are the following: The group membership layer, ....

....the messages broadcast to the current group members. This semantics is called view synchrony (see Section 1) The upper layer provides atomic broadcast : it ensures that messages are delivered in the same order by all processes. Atomic broadcast is implemented using the view synchrony layer [8]. Phoenix. The Phoenix architecture [25] is a variation of the Isis architecture (Fig 2) The basic layer solves the consensus problem [10] Membership (primary partition) and view synchrony are provided by the same layer: both the membership problem and view synchrony are solved using the ....

[Article contains additional citation context not shown here]

K. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Trans. on Computer Systems, 9(3):272--314, August 1991.

....decides its own proposal and sends the decision (using reliable broadcast) to all other processes. The other processes decide upon receiving the decision message. 4. 2 Fixed sequencer uniform atomic broadcast algorithm The second uniform atomic broadcast algorithm is based on a fixed sequencer [19]. It uses a group membership service for reconfiguration in case of a crash. We shall refer to it as the GM atomic broadcast algorithm, or simply as the GM algorithm. We describe here the uniform version of the algorithm. In the GM algorithm, one of the processes takes the role of sequencer. When ....

K. Birman, A. Schiper, and P. Stephenson, "Lightweight causal and atomic group multicast," ACM Transactions on Computer Systems, vol. 9, pp. 272--314, Aug. 1991.

....sends back the reply to the client. Making the replication protocol correct when the controllers may crash and network messages may be lost is more difficult. However, much of the difficulty can be overcome by devising the replication protocol using the View Synchronous Communication abstraction [1, 2, 10, 11]. Figure 4 shows the architecture of this approach: a straightforward primary backup protocol (upper box) based on a VSC protocol (lower box) The VSC layer provides a Group Membership Service and VSC communication primitives that we describe next. Primary backup replication layer (Sect. 3.4) ....

K. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Trans. on Computer Systems, 9(3):272--314, August 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Trans. on Computer Systems, 9(3):272-314, Aug. 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson, "Lightweight causal and atomic group multicast", ACM Trans. Compt. Syst. Vol. 9, No. 3, pp. 272-314, Aug. 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Transactions on Computer Systems, 9(3):272--314, 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Transactions on Computer Systems, 9(3):272--314, August 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Transactions on Computer Systems, 9(3):272--314, Aug. 1991.

No context found.

Birman, K., Schiper, A., and Stephenson, P. 1991. Lightweight causal and atomic group multicast. ACM Transactions on Computer Systems 9, 3 (Aug.), 272--314.

No context found.

Kenneth Birman, Andre Schiper, and Pat Stephenson. Lightweight causal and atomic group multicast. ACM Transactions on Computer Systems, 9(3):272-314, August 1991.

No context found.

Kenneth Birman, Andr'eSchiper, and Pat Stephenson. Lightweight causal and atomic group multicast. ACM Trans. on Computer Systems, 9(3):272--314, Aug. 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson, "Lightweight Causal and Atomic Group Multicast," ACM Trans. Computer Systems, vol. 9, no. 3, pp. 272-314, Aug. 1991.

No context found.

Birman K., Schiper A., and Stephenson P.,"Lightweight Causal and Atomic Group Multicast", ACM Transactions On Computer Systems, Vol. 9, No. 3, August 1991, pp. 272-314.

No context found.

Birman, K., Schiper, A. and Stephenson, P. (1991) Lightweight causal and atomic group multicast. ACM Trans. Comput. Syst., 9, 272--314.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Trans. Comput. Syst., 9(3):272--314, 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson, "Lightweight Causal and Atomic Group Multicast," ACM Transactions on Computer Systems, vol. 9, no. 3, Aug 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Transactions on Computer Systems, 9(3), August 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Transactions on Computer Systems, 9(3):272--314, August 1991.

No context found.

K. Birman, A. Schiper, and P. Stephenson,"Lightweight Causal and Atomic Group Multicast", ACM Transactions On Computer Systems, Vol. 9, No. 3, August 1991, pp. 272-314.

No context found.

K. Birman, A. Schiper, and P. Stephenson, Lightweight Causal and Atomic Group Multicast, ACM Trans. Computer Systems 9, 3(Aug. 1991), 272--314.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Trans. Comput. Syst., 9(3):272-314, 1991.

No context found.

Birman, K., Schiper, A., and Stephenson, P. Lightweight causal and atomic group multicast. ACM Trans. Comput. Systems 9, 3 (1991), 272--314.

No context found.

K. Birman, A. Schiper, and P. Stephenson. Lightweight causal and atomic group multicast. ACM Trans. Comput. Syst., 9(3):272-314, 1991.

No context found.

Birman, K. P., Schiper, A. and Stephenson, P. "Lightweight Causal and Atomic Group Multicast". Trans. Computer Systems 9, 3 (August 1991), pp.272-314.

No context found.

K. P. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Transactions on Computer Systems, 9(3):272--314, 1991.

No context found.

Birman, K., Schiper, A., and Stephenson, P. Lightweight causal and atomic group multicast. ACM Trans. Comput. Syst. 9, 3 (1991), 272--314.

No context found.

Kenneth Birman, Andr e Schiper, and Pat Stephenson. Lightweight causal and atomic group multicast. ACM Transactions on Computer Systems, 9(3):272--314, August 1991.

No context found.

K. P. Birman, A. Schiper, and P. Stephenson. Lightweight Causal and Atomic Group Multicast. ACM Trans. on Computer Systems, 9(3):272--314, Aug. 1991.

No context found.

K. Birman, A. Schiper and P. Stephenson, Lightweight Causal and Atomic Group Multicast, ACM Transactions on Computer Systems, Vol. 9, No. 3, pp. 271-314, 1991.

No context found.

K. P. Birman, A. Schiper and P. Stephenson, `Lightweight causal and atomic group multicast', ACM TOCS, 9, (3), 272--314 (1991).

No context found.

K. Birman, A. Schiper and P. Stephenson, `Lightweight causal and atomic group multicast', ACM Trans. Computer Systems, 9, 272--314 (1991).

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