N. Krishnakumar and A.J. Bernstein, "Bounded Ignorance in Replicated Systems," in Proc. ACM-PODS'91, Denver, CO, May 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....that are allowed to perform a certain action, such as issuing broadcast messages. In such a case, the system may restrict the number of broadcasting processes so as to control the level of congestion. Another application is in the context of replicated databases that allow bounded ignorance [11], i.e. when transactions may specify that they do not need to be aware of the k most recent updates to the database. Here also, instead of the traditional database system that uses distributed mutual exclusion to ensure one update to the replicated data at any time, several updates may be ....

N. Krishnakumar and A.J. Bernstein, "Bounded Ignorance in Replicated Systems," Proc. 10th ACM Symp. Principles of Database Systems, pp. 63--74, May 1991.

....to improve the performance in a replicated database is to relax the serializability requirement. Works on quasi copies ( Alonso et al. 1988; Alonso et al. 1990; Barbara and Garcia Molina 1990] lazy replication ( Ladin et al. 1988; Ladin et al. 1992; Ladin et al. 1990] and bounded ignorance ([Krishnakumar and Bernstein 1991]) fall in this category. In contrast, as we show in section 3, the adaptive replication algorithms that we propose here can be combined with a concurrency control algorithm to preserve 1 copy serializability. Recently, a few works that address the problem of dynamic (vs. static) data replication ....

Krishnakumar, N. and Bernstein, A. 1991. Bounded ignorance in replicated systems. In Proceedings of the ACM Principles of Database Systems.

....any two consecutive steps of a transaction. A schedule is correct if it is equivalent to a schedule in which steps are executed serially, and between any two consecutive steps of a transaction, only steps of those transactions that are permitted to interleave appear in the schedule. In [PL91, KB91, WA92] the authors propose schemes that exploit the semantics of applications in order to bound the inconsistency due to non serializable executions. The approach we adopt in this paper enables the semantics of transactions to be exploited, and is similar to the one used in [GM83, FO89] ....

N. Krishnakumar and A. J. Bernstein. Bounded ignorance in replicated systems. In Proceedings of the tenth ACM SIGACT-SIGMOD-SIGART Symposium on Principles of Database Systems, Denver, pages 63--74, May 1991.

....inconsistent. As another example, IBM Data Replicator uses a pull strategy whereby a client will not see its own updates unless it requests them. 2.2 . lazy solutions . On the research side, lazy replication has been studied using very different approaches like weak consistency models [PL91, KB91, GN95] economic paradigms [SAS 96] or epidemic strategies [AES97] More recent work has explored lazy strategies that still provide consistency. Thus, Chundi et al. CRR96] have shown that in lazy primary copy schemes, serializability cannot be guaranteed without restricting the placement of ....

N. Krishnakumar and A.J. Bernstein. Bounded ignorance in replicated systems. In Proc. of PODS, 1991.

....node can make such decisions to grow or shrink the replication scheme. The third approach is to assume static replication but to relax the serializability criterion. Papers on quasi copies ( ABM90] BaM90] lazy replication ( LLS88] epsilon serializability ( PuL91] and bounded ignorance ([KrB91]) fall in the category. The serializability relaxation can be done along two dimensions of time and space( ShR90] While divergence may be acceptable in some cases (e.g. stock market data) strict consistency is a must in many distributed applications like banking databases. Our scheme is geared ....

N. Krishnakumar and Arthur Bernstein, "Bounded Ignorance in Replicated Systems, " ACM PODS (1991).

....that are allowed to perform a certain action, such as issuing broadcast messages. In such a case, the system may restrict the number of broadcasting processes so as to control the level of congestion. Another application is in the context of replicated databases that allow bounded ignorance [11], i.e. when transactions may specify that they do not need to be aware of the k most recent updates to the database. Here also, instead of the traditional database system that uses distributed mutual exclusion to ensure one update to the replicated data at any time, several updates may be ....

N. Krishnakumar and A.J. Bernstein, "Bounded Ignorance in Replicated Systems," Proc. 10th ACM Symp. Principles of Database Systems, pp. 63--74, May 1991.

.... DATABASES3 accepted on a subset of the sites, or transactions that are known not to lead to inconsistency are the only ones allowed to run [3] These protocols suffer from unrealistically strong assumptions [34] therefore, a lot of work has been done to allow some more flexibility [4] 9] [25]. A noticeable example of this is the work on epsilon serializability [32] 33] Updates are allowed to propagate through the system asynchronously; eventual consistency of the system is an asymptotical property. A whole family of methods, each one different in the level of asynchrony, can be ....

N. Krishnakumar and A.J. Bernstein, "Bounded Ignorance in Replicated Systems," in Proc. ACM-PODS'91, Denver, CO, May 1991.

.... by either the two phase locking rule [EGLT76] or the time stamp ordering rule [Ree78] In the context of the above query update situation, where we do not require an absolutely shortest alternative route, it is possible to relax the correctness condition of serializability to bounded ignorance [KB91] or bounded inconsistency [WA92] This allows a greater degree of concurrency and better performance. A variation of such relaxed correctness condition is used in our system. We would like to bound the error E i in the estimated routing time for each transaction i (or query in our system) This ....

N. Krishnakumar and A. J. Bernstein. Bounded Ignorance in Replicated Systems. In Proceedings of the 10th ACM Symposium on Principles of Database Systems, pages 63--74, May 1991.

....scheme constant during runtime. Another approach to improve the performance in a replicated distributed database, which also assumes a static replication scheme, is to relax the serializability requirement. Works on quasi copies ( 13, 14, 15] lazy replication (in [16] and bounded ignorance ([17]) fall in this category. In contrast, our approach preserves one copy serializability. In the theoretical computer science community there has been work on online algorithms (e.g. 1] particularly for paging (e.g. 18] searching (e.g. 18] and caching (e.g. 19] These works are similar in ....

N. Krishnakumar and A. Bernstein, Bounded ignorance in replicated systems, Proc. of ACMPODS '91.

....updates, but is inherently discrete and batch processing oriented. Even some banking applications do not need atomic updates of replicas. Therefore, several protocols have been developed for updating replicated data without the requirement of atomic and synchronous update to each and every replica [1, 8, 11, 13, 16, 17, 20, 21]. Some of these protocols work by transforming global constraints on the data into local constraints that should hold on the replicas; each replica may then independently be updated as long as its local constraints are satisfied [5] Other protocols recover from violations of global constraints by ....

N. Krishnakumar and A.J. Bernstein, "Bounded ignorance in replicated systems, " in Proc. ACM-PODS'91, Denver, CO, May 1991.

....[DF] for a survey) In contrast, in this paper we discussed dynamic replication. Another approach to improve the performance in a replicated distributed database is to relax the serializability requirement. Works on quasi copies ( ABG1, ABG2, BG] lazy replication ( LLS] and bounded ignorance ([KB]) fall in this category. These works also assume a static replication scheme. In contrast, our approach preserves one copy serializability, since the DDA algorithm is read one write all (although the meaning of all changes dynamically) In the theoretical computer science community there has ....

N. Krishnakumar and A. Bernstein, "Bounded ignorance in replicated systems," Proc. of ACM-PODS '91.

.... either updates are accepted on a subset of the sites, or transactions that are known not to lead to inconsistency are the only ones allowed to run [3] These protocols suffer from unrealistically strong assumptions [32] therefore, a lot of work has been done to allow some more flexibility [4, 9, 23]. A noticeable example of this is the work on epsilonserializability [30, 31] Updates are allowed to propagate through the system asynchronously; eventual consistency of the system is an asymptotical property. A whole family of methods, each one different in the level of asynchrony, can be ....

N. Krishnakumar and A.J. Bernstein, "Bounded ignorance in replicated systems," in Proc. ACM-PODS'91, Denver, CO, May 1991.

....words: distributed systems, causality, logical time, matrix time, fault tolerance. 1 Introduction Matrix clocks have been introduced in the context of asynchronous distributed systems. They have nice properties that can be used to design distributed database protocols and fault tolerant protocols [WB84, KB91]. Unfortunately, they are costly to implement: in a distributed system that consists of n sites, the naive algorithm to compute the matrix clock on the fly requires that an n Theta n matrix of integers be stored at each site and tagged onto each message: if the number of sites is large, it is ....

....crash failure semantics such that at most k Gamma 1 faults may occur simultaneously. For example, we plan to use k matrix clocks to implement a crash resilient event logging facility for the CDB distributed debugger [Rug94] There is a similarity between k matrix clocks and k bounded ignorance [KB91]. However, whereas k bounded ignorance is a distributed database technique to guarantee that a given transaction cannot be ignorant of more than k (causally) preceding transactions, k matrix clocks guarantee that no more than k computation sites can be ignorant of a (causally) preceding event. ....

Krishnakumar and Bernstein. Bounded ignorance in replicated systems. In Proc. ACM symp. on Principles of Database Systems, 1991.

....system that are allowed to perform a certain action, such as issuing broadcast messages. In such a case, the system may restrict the number of broadcasting processes so as to control the level of congestion. Another application is in the context of replicated databases that allow bounded ignorance [10], i.e. when transactions may specify that they do not need to be aware of the k most recent updates to the database. Here also, instead of the traditional database system that uses distributed mutual exclusion to ensure one update to the replicated data at any time, several updates may be ....

N. Krishnakumar and A. J. Bernstein. Bounded Ignorance in Replicated Systems. In Proc. of 10th ACM Symp. on Prin. of Database Sys., pp. 63--74, May 1991.

....yet capture enough information about the system to be useful. Existing algorithms to record distributed snapshots [1, 4, 8, 13] show that they are intended for systems with message delivery semantics no stronger than FIFO. On the other hand, there exist a number of recent distributed applications [2, 3, 5, 6, 7], wherein messages are sent and delivered according to a message ordering protocol that implements a communication abstraction stronger than FIFO message delivery. A natural question to ask therefore, is whether it is any easier to record distributed snapshots in such systems that support more ....

N. Krishnakumar and A. J. Bernstein. Bounded ignorance in replicated systems. Proceedings of the 10 th ACM Symposium on Principles of Database Systems, 1991.

....work has also been done in constructing weak consistency models. These models typically place bounds on one or more divergence parameters. For example, some systems place bounds on time (temporal divergence control) AGRA93] or the number of update transactions (value based divergence control) [KRIS91]. Finally, other systems focus on flexible specification of divergence parameters. Both quasi copies [ALON90] and epsilon serializability [PU91] permit value based and temporal divergence control on the underlying data. Bayou [TERR94] takes a slightly different approach, providing various kinds of ....

N. Krishnakumar and A. J. Bernstein. Bounded ignorance in replicated systems. Proc. 10th ACM SIGACT-SIGMOD Conf. on Principles of Database Sys., pages 63--74, May 1991.

....the interleaved execution of operations on an object is weaker than the one derived from the commutativity of operations, and hence it increases the concurrency in the system. The drawback of using this approach is that the amount of inconsistency introduced to a transaction may be unbounded. In [17], the notion of N ignorant transaction, which is a transaction that may be ignorant of the results of at most N prior transactions, is introduced to increase concurrency at the expense of consistency. Although the number of transactions that are ignored is bounded, the inconsistency introduced by ....

N. Krishnakumar and A. J. Bernstein. Bounded Ignorance in Replicated Systems. In Proceedings of the Tenth ACM Symposium on Principles of Database Systems, pages 63--74, 1991.

No context found.

N. Krishnakumar and A.J. Bernstein, "Bounded Ignorance in Replicated Systems," in Proc. ACM-PODS'91, Denver, CO, May 1991.

No context found.

N. Krishnakumar and A.J. Bernstein. Bounded Ignorance in Replicated Systems. In Proceedings of the 1991 ACM Symposium on principles of Database Systems, pages 63-74, May 1991,

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