P.A. Bernstein, and N. Goodman. An algorithm for concurrency control and recovery in replicated distributed databases. ACM Trans. on Database Sys., 9(4):596-615, December 1984.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....as this is the main reason for replicating data. The consistency condition referred to is one copy equivalence [7] Coupled with serializable executions that many databases provide, the correctness criteria for replica control protocols in transactional environments is one copy serializability [6]. Among the replica control protocols that can achieve one copy serializability when used in conjunction with appropriate mechanisms for concurrency control (e.g. two phase locking) and atomic commitment (e.g. two phase commit) we identify the read one write all approach (ROWA) 4] the ....

....Let a read quorum group be V , and a write quorum group be V [ U . Any two write quorum groups intersect, and any read quorum group intersects any other write quorum group. The quorum intersection property is satis ed and reads and writes are serializable. In particular, one copy serializability [6] is guaranteed. On the left side of Figure 1 we illustrate read and write quorum groups for a 3 space quorum set. For clarity of exposition we will assume hereafter that the extension of the replica space is the same along all dimen The condition holds only when d is even and k = d=2. If d is ....

P. A. Bernstein and N. Goodman. An algorithm for concurrency control and recovery in replicated distributed databases. ACM Transactions on Database Systems, 9(4):596-615, December 1984.

....for the copies of a data item. In our replicated database system model, concurrency control is provided by any of the concurrency control protocols presented in the following sections, and mutual consistency of replicated data is achieved by using the read one, write all available scheme [8]. The reason for selecting this replica control scheme is that alternatives like quorum based approaches have the major drawback of turning read operations into multisite operations, even for local data [9, 12] Based on the read one, write all available approach, a read operation on a data ....

P.A.Bernstein, N.Goodman, "An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases", ACM Transactions on Database Systems, vol.9, pp.596-615, 1984. 28

....will be able to obtain service. However, this arrangement has serious maintainability and transparency problems. First, the client must treat replicated files in a different way from non replicated files. Second, if the replication algorithm is changed, perhaps to the available copies algorithm [2], then every client must be found and modified to reflect that change. In a wide area distributed system this is infeasible. The obvious solution to this problem is to encapsulate the access data in a clerk object; when a client object needs to read a file, it invokes a clerk, which then executes ....

Bernstein, P. A. and Goodman, N. "An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases". Trans. Database Systems 9, 4 (December 1984), pp.596-615.

....a view change has no effect on what client ordered updates are known in the new view. Instead, these continue to be propagated by gossip just like they were in the system of Section 3. If there are many server ordered updates, the primary may become a bottleneck. In this case, a voting method [11, 2, 13] could be used, but this has the disadvantage that it will be blocking (i.e. if the client becomes disconnected from the service during phase 1, no more server ordered updates could be performed until the client recovered) Blocking can be avoided by using a three phase protocol [32] but then ....

....the needed information will be contained in the specifications of the modules with which the client interacts. 6. Related Work Our work builds on numerous previous results in the area of highly available distributed systems and algorithms, including general replication techniques such as voting [11, 13, 2] and the primary copy method [1, 27, 26] Our work is also related to gossip schemes [12, 10, 34] In this section, we focus on the most closely related work, namely providing high availability for applications where operations need not be ordered identically at all replicas. In this light, we ....

[Article contains additional citation context not shown here]

Bernstein, P. A., Goodman, N. An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases. AC M Transactions on database Systems 9(4):596-615, December, 1984.

....1. Introduction Data replication techniques have been used to increase reliability and availability of data in distributed computing systems (DCS) in the presence of failures [14] 17] Numerous algorithms have been proposed for maintaining consistency among replicated data objects [6], 7] 16] 21] Included in these is a class of votingbased algorithms which tolerate both site and communication link failures [4] 8] 15] 19] 20] Recently, regeneration of failed copies of data objects has been proposed to improve availability and reliability of data [1] 3] 10] ....

P.A. Bernstein and N. Goodman, "An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases," ACM Transactions on Database Systems, Vol. 9, No. 4, pages 596-615, December 1984.

....data replication is to improve performance. With many copies of each data item being available, a user transaction is more likely to find the data it needs nearby. However, these benefits are offset by the cost of maintaining data consistency. Synchronization protocols, such as those mentioned in [4, 5, 10, 13, 2] are needed to coordinate the operations on the replicas. This increases the cost of executing operations in replicated databases. One of the proposed protocols for the above problem is the hierarchical quorum consensus protocol [12] In this paper, we generalize the hierarchical quorum consensus ....

P. A. Bernstein, N. Goodman, An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases, ACM Transactions on Database Systems, Vol.9, No.4, 1984, pages 596-615. 18

....an operation depend on the number of replicas which have to be accessed. In order to provide low communication costs, the number of replicas should be kept as small as possible. In the literature, there exist several replication algorithms aimed at different goals. The Read One WriteAll Algorithm [3] is a straightforward approach that has the advantage of minimizing the cost and maximizing the availability of the read operation at the same time. Unfortunately, this strategy has severe drawbacks for the write operation, which has maximal communication cost, i.e. all replicas must be written. ....

P. A. Bernstein and N. Goodman. An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases. ACM Transactions on Distributed Systems, 9(4), 1984.

....y The work of this author was done while a Visiting Scientist at IBM Almaden Research Center. A first class of protocols makes the assumption that network partitions are either unlikely or unlikely to occasion conflicting updates. The best known of them are the available copy protocol (AC) [2, 7], the regeneration algorithm [17] and the Coda replication control protocol [18] The second class of protocols take the approach that data consistency is much more important than data availability. These protocols rely on quorums to provide mutual exclusion and prevent conflicting updates. As a ....

....the attention that it deserves. As we will see, the results of this neglect have been replication control protocols with bloated metadata and complex procedures for ascertaining which replicas are up to date. We present a new implementation of Bernstein andGoodman s available copy protocol [2]. Our new protocol maintains for each replica a cohort set that is updated any time a failure is detected or a replica residing on a site that failed 400 is repaired. By requiring that all changes in the cohort set involve all sites in the new cohort set, we guarantee that all replicas sharing ....

[Article contains additional citation context not shown here]

P. A. Bernstein and N. Goodman, "An algorithm for concurrency control and recovery in replicated distributed databases," ACM Transactions on Database Systems, Vol. 9, No. 4 (1984), pp. 596--615.

....replicas receive them in the same order, b) broadcasting these writes to all available replicas, and (c) requiring that replicas residing on nodes recovering from a failure remain unavailable or comatose until they are brought up to date. This available copies protocol has two major advantages [2, 17]. First, read requests never need to access more than one available replica because all available replicas are guaranteed to be up to date. Second, the replicated data can be accessed as long as there is at least one available replica. The situation is quite different when network partitions must ....

Bernstein, P.A., Goodman, N.: An Algorithm for concurrency control and recovery in replicated distributed databases. ACM Trans. on Database Systems, 9, 4 (1984) 596--615.

....number of copies in the write quorum plus the number of copies in the read quorum exceeds the total number of copies for the object. The overlap itself guarantees that any R(X) will return the result from the last W (X) The first ROWAA algorithm (Available Copies) was proposed by Bernstein et al. [4]. The idea is to write on all copies that are accessible and therefore allow reading of any accessible copy. The copies that are down during an update must be brought up to date, however, during the node recovery time. Compared to quorum consensus, ROWAA offers a faster R(X) just one r(x) instead ....

P.A. Bernstein and N. Goodman. An algorithm for concurrency control and recovery in replicated distributed databases. ACM Transactions on Database Systems, 9(4):596-- 615, December 1984.

....network traffic and incurs no significant storage cost. Finally, regenerable volatile witnesses can be combined with two tier voting to provide a high level of fault tolerance with as few as two replicas in stable storage. Unlike other replication control protocols that require only two replicas [BeGo84, AgEl88, Pari86, ReTa88, Pari89], our novel technique requires only two sites providing stable storage, operates correctly in the presence of communication failures, and does not require any knowledge of the network topology. The remainder of this paper is organized as follows: Section two introduces regenerable volatile ....

P. A. Bernstein and N. Goodman, "An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases," ACM Transactions on Database Systems, 9, 4 (1984), pp. 596-615.

....than replicas. We can view these witnesses as the distributed equivalent of a single very reliable witness that would know at any time which replicas are current and would enforce mutual exclusion. We would thereby achieve a very close approximation of the optimistic available copy protocol [3], which does not guarantee the consistency of the replicated data in the presence of network partitions but provides much higher data availabilities than voting protocols. V. CONCLUSIONS Voting protocols have been extremely popular during the last ten years due to their robustness and their ....

P. A. Bernstein and N. Goodman, "An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases," ACM TODS, Vol. 9, No. 4 (1984), pp. 596--615.

....concerned. Therefore, the interleaved execution of transactions on a replicated database should be equivalent to a serial execution of these transactions on a single copy database. Transaction processing in replicated database systems has been studied by a number of researchers (see for example [AE90, AE92, BHG87, BG84, BL93, PL91] and the references contained therein) Several protocols 1 Research Supported by NSF Grant CCR 90 06396. Email addresses: fparu, djr, ravig cs.albany.edu achieving replica consistency, such as two phase commit and quorum consensus, have been proposed in the literature [BL93, BHG87, Gi79, ....

P. A. Bernstein and N. Goodman, "An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases," ACM TODS, Vol. 9, No. 4, Dec. 1984, pp 596-615.

....Ideally, the interleaved execution of transactions in a replicated database system should be equivalent to a serial execution of these transactions on a single copy database. Transaction processing in replicated database systems has been studied by a number of researchers (see for example [AE90, AE92, BHG87, BG84, PL91] and the references contained therein) Several protocols achieving replica consistency, such as two phase commit and quorum consensus, have been proposed in the literature [BHG87, Gi79, Th79] These protocols allow a transaction to access a replicated data item at a site only if the transaction ....

P. A. Bernstein and N. Goodman, "An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases," ACM TODS, Vol. 9, No. 4, Dec. 1984, pp 596-615.

....van Renesse Andrew S. Tanenbaum Dept. of Computer Science Vrije Universiteit The Netherlands ABSTRACT Data replication is a technique for increasing the availability of data. Two popular algorithms for maintaining consistency among the replicas are Weighted Voting [1] and Available Copies [2] In recent papers [3] it has been shown that under common circumstances Available Copies (AC) performs better than Weighted Voting (WV) However, the issue of network partitioning due to gateway crashes is ignored in AC. We present an improvement of WV that, if configured accordingly, performs as ....

....data look like a single object, even under concurrent access. Users should always see the most recent version. This prerequisite is called serial consistency or one copy serializability . Two popular algorithms for achieving serial consistency are Weighted Voting [1] and Available Copies [2] Weighted Voting (WV) trades off read availability for write availability, and may need more than one copy available to perform an operation. Available Copies (AC) works as long as there is at least one available copy, but works incorrectly in the presence of communication errors or network ....

P. A. Bernstein and N. Goodman, "An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases," ACM Trans. on Database Systems, vol. 9, no. 4, pp. 596-615, December 1984.

....that has been forced to a majority of cohorts and the new view contains every event that is known by this cohort. 2. 2 Available Copies and Regeneration Methods Bernstein and Goodman studied the naive read one write all available consistency control protocol in a network that cannot partition [BG84] They show that there exist sequences of node failures that can lead to inconsistency when the read one write all available method is used. They have extended this method so it operates correctly as long as network does not partition. The available copies method uses a directory locking scheme ....

....the available copies method will guarantee correctness. If the boot service becomes unavailable for some reason, the voting with ghosts protocol degrades to the weighted voting method. If the network consists of a single segment, the protocol operates identically to the available copies method [BG84] It is also shown that the vote assignment will have significant impact on the data availability and the availability achieved using voting with ghosts is higher than weighted voting. The voting with bystander scheme [P 89] assumes a network consisting of segments that are immune to partial ....

[Article contains additional citation context not shown here]

P.A. Bernstein and N. Goodman. An algorithm for concurrency control and recovery in replicated distributed databases. ACM Transactions on Database Systems, 9(4):596--615, 1984.

....d, a quorum must be collected. All nodes r for which AC t (d; r) 6= 0 are asked to vote in case of a read or write request. The quorum size qu t (d) that must be reached at a time t, is 1 qu t (d) trunc P r2R AC t (d; r) 2 1 (3) Our main goal is the consistency of all the data files D [5, 6, 7]. A crash renders the node s data file temporarily or permanently inaccessible, while communication link failures result in messages to be lost. Therefore, we assume the existence of stable storage devices [32] and use a two phase locking protocol within a transaction oriented scheme [44] ....

P.A. Bernstein and N. Goodman. An algorithm for concurrency control and recovery in replicated distributed databases. ACM Transactions on Database Systems, 9(4):596--615, December 1984.

No context found.

P.A. Bernstein, and N. Goodman. An algorithm for concurrency control and recovery in replicated distributed databases. ACM Trans. on Database Sys., 9(4):596-615, December 1984.

No context found.

P. A. Bernstein and N. Goodman. An algorithm for concurrency control and recovery in replicated distributed databases. ACM Trans. on Database Sys., 9(4):596--615, Dec. 1984.

No context found.

P. A. Bernstein and N. Goodman. An algorithm for concurrency control and recovery in replicated distributed databases. ACM Transactions on Database Systems, 9(4):596--615, December 1984.

No context found.

Bernstein, P. A. and Goodman, N. "An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases". Trans. Database Systems 9, 4 (December 1984), pp.596-615.

No context found.

P. A. Bernstein and N. Goodman, `An algorithm for concurrency control and recovery in replicated distributed databases', ACM Trans. Database Systems, 9, (4), 596--615 (1984).

No context found.

P. A. Bernstein and N. Goodman, `An algorithm for concurrency control and recovery in replicated distributed databases', ACM Trans. Database Systems, 9, (4), 596--615 (1984).

No context found.

P.A. Bernstein and N. Goodman. An algorithm for concurrency control and recovery in replicated distributed databases. ACM Transactions on Database Systems, 9(4):596-615, December 1984.

No context found.

BeGo84 P.A. Bernstein and N. Goodman, "An Algorithm for Concurrency Control and Recovery in Replicated Distributed Databases," ACM Trans. on Database Systems, 9 (4) (1984) 596-615.

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