A. Fekete, D. Gupta, V. Luchangco, N. Lynch, and A. Shvartsman. Eventually-serializable data services. Theoretical Computer Science, 220(Special issue on Distributed Algorithms):113--156, 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... an Eventually Serializable Data Service as a Distributed System Building Block 1999 Abstract This work presents an implementation of a distributed system building block that is formally speci ed as the Eventually Serializable Data Service (ESDS) [7] proposed by Fekete et al. ESDS deals with replicated objects that allow the users of the service to relax consistency requirements in return for improved responsiveness, while providing guarantees of eventual consistency of the replicated data. The ESDS paper [7] includes a formal service ....

....Data Service (ESDS) 7] proposed by Fekete et al. ESDS deals with replicated objects that allow the users of the service to relax consistency requirements in return for improved responsiveness, while providing guarantees of eventual consistency of the replicated data. The ESDS paper [7] includes a formal service speci cation and an abstract algorithm implementing the service. The algorithm is given in terms of I O automata of Lynch and Tuttle [15] An important consideration in formulating ESDS was that it could be employed in building real systems. The work described here ....

[Article contains additional citation context not shown here]

A. Fekete, D. Gupta, V. Luchangco, N. Lynch, and A. Shvartsman. EventuallySerializable Data Services. Principles of Distributed Systems 1996, pp. 300-310.

....how it can be converted to atomicity. In Section 5. 2, I extend the proofs to the dynamic case moving nodes, and finally in Section 6 I discuss the algorithm s 1Speaking informally, this extremely weak property looks to capture the benefits of both the eventually serializable property in [7] and the weak sequential locking in [12] Umbra Refltor Node Figure 2: Anatomy of a persistent node. The innermost circle is the node proper (red) where data can be read or written. Every particle within middle circle is in the reflector, which holds data and calculates which direction the ....

Alan Fekete, David Gupta, Victor Luchangco, Nancy Lynch, and Alex Shvartsman. EventuallySerializable Data Services. Theoretical Computer Science, 220(1):113-156, June 1999. Special Issue on Distributed Algorithms.

....all the updates to all the replicas. This naive algorithm has an obvious downside, however: it becomes slower and consumes more disk space as it accumulates more updates. Timestamp vector is a technique used in pull based systems to minimize the number of updates exchanged between replicas [60, 25, 61, 1, 45, 20]. Each replica keeps an on disk timestamp vector TV , an N element array of timestamps, which summarizes the state of the replica. Here, N is the total number of replicas in the system and the timestamp is any number that increases monotonically a logical clock [46] a wall clock [50, 51] or ....

Alan Fekete, David Gupta, Victor Luchangco, Nancy Lynch, and Alex Shvartsman. Eventually Serializable Data Services. In 15th ACM Symp. on Princ. of Distr. Computing (PODC), Philadelphia, PA, May 1996.

....green yellow green Figure 3.4 The MQs after after recovery. As explained in Amir et al. 1993; Keidar, 1994, it is possible to provide the application with red messages if weak consistency guarantees are required. For example, eventually serializable data services (e.g. Pu and Leff, 1991; Fekete et al. 1996; Amir et al. 1993) deliver messages to the application before they are totally ordered. Later, the application is notified when the message becomes stable (green in our terminology) Messages become stable at the same order at all processes. The advantage of using COReL for such applications is ....

....becomes stable (green in our terminology) Messages become stable at the same order at all processes. The advantage of using COReL for such applications is that with COReL messages become stable even whenever a majority is connected, while with the implementations presented in Pu and Leff, 1991; Fekete et al. 1996; Amir et al. 1993, messages may become stable only after they are received by all the processes in the system. 5.3 NOTATION We use the following notation: MQ p is the MQ of process p. Prefix (MQ p ; m) is the prefix of MQ p ending at message m. Green(MQ p ) is the green prefix of MQ p . We ....

Fekete, A., Gupta, D., Luchangco, V., Lynch, N., and Shvartsman, A. (1996). Eventually-serializable data services. In 15th ACM Symposium on Principles of Distributed Computing (PODC), pages 300--309.

....algorithm can distribute all the updates to all the replicas. It has an obvious downside, however: it becomes slower and consumes more disk space as it accumulates more updates. Timestamp vector is a technique used in pull based systems to minimize the number of updates exchanged between replicas [57, 23, 58, 1, 43, 19]. Each replica keeps an on disk timestamp vector TV , an N element array of timestamps, that summarizes the state of the replica. Here, N is the total number of replicas and the timestamp is any number that increases monotonically a logical clock [44] a wall clock [48, 49] or a counter that ....

Alan Fekete, David Gupta, Victor Luchangco, Nancy Lynch, and Alex Shvartsman. Eventually serializable data services. In ACM Symp. on Princ. of Distr. Computing, May 1996.

....Sequential consistency [11] guaranteed by systems such as Orca [12] allows operations to be reordered as long as they remain consistent with the view of isolated clients. Other systems provide even weaker guarantees to the clients [13, 14, 15] to get better performance. Fekete et al. [1] defined a highly available eventually serializable data service (ESDS) They specified general conditions for such a service, and presented an algorithm based on lazy replication, in which operations received by each replica are gossiped in the background. Responses to operations may be ....

....report specifies the semantics of the failure models and the desired behavior of working replicas in the face of failures. To compensate for the lack of immediate consistency guarantees, ESDS provides a mechanism for specifying dependencies between operations explicitly. The abstract algorithm in [1] uses prev sets to identify the dependencies. Each operation x submitted to the system includes a prev set of operations that must be applied to the state before x. However, it is impractical to require users of an ESDS based system to specify large dependency sets. Furthermore, prev sets are ....

[Article contains additional citation context not shown here]

A. Fekete, D. Gupta, V. Luchangco, N. Lynch, and A. Shvartsman. Eventually-Serializable Data Services. PODC 1996, pp. 300-310.

.... and Computer Science August 21, 1997 In Partial Fulfillment of the Requirements for the Degrees of Bachelor of Science in Computer Science and Engineering and Master of Engineering in Electrical Engineering and Computer Science ABSTRACT This thesis builds on the work of Fekete et al. [1], who defined an eventually serializable data service (ESDS) and an abstract algorithm for it. ESDS allows its users to relax consistency requirements in return for improved responsiveness, while providing guarantees of eventual consistency of the replicated data. An important consideration in ....

....Specification of distributed systems building blocks and development of supporting algorithms is one of the main research areas at the Theory of Distributed Systems group (TDS) at the MIT Laboratory for Computer Science. TDS recently defined a flexible eventuallyserializable data service (ESDS) [1]. The definition includes a formal specification of the data service and an abstract distributed algorithm that implements the service. ESDS relaxes consistency guarantees provided by serializable distributed data services to improve system efficiency and availability. It also provides provable ....

[Article contains additional citation context not shown here]

A. Fekete, D. Gupta, V. Luchangco, N. Lynch, and A. Shvartsman. EventuallySerializable Data Services. PODC 1996, pp. 300-310.

No context found.

A. Fekete, D. Gupta, V. Luchangco, N. Lynch, and A. Shvartsman. Eventually-serializable data services. Theoretical Computer Science, special issue on Distributed Algorithms, 220, 1999.

....Service Oleg M. Cheiner Alex A. Shvartsman Replication is used in distributed systems to improve availability and to increase throughput. The disadvantage of replication is the additional effort required to maintain consistency among replicas when serializing client operations. Fekete et al. [1] proposed the Eventually Serializable Data Service (ESDS) that maintains replicated objects and allows the clients of the service to relax consistency requirements in return for improved responsiveness. ESDS guarantees eventual consistency of the replicated data, while allowing the clients to ....

....with guaranteed consistent responses and non strict operations that return responses consistent with some ordering of operations. ESDS builds on the work of Ladin et al. 3] who defined a data service with relaxed consistency and presented an algorithm based on lazy replication. The ESDS paper [1] includes a service specification and an abstract distributed algorithm; both are given in terms of I O automata of Lynch and Tuttle [2] The abstract algorithm is shown to implement the service specification, in the sense of trace inclusion. The work summarized here aims to extend the formal ....

[Article contains additional citation context not shown here]

A. Fekete, D. Gupta, V. Luchangco, N. Lynch, and A. Shvartsman. Eventually-Serializable Data Services. PODC 1996, pp. 300-310.

....The Obj Data Type The application manages deterministic data objects whose serial behavior is specified by some data type, Obj. The Obj data type defines possible states of the objects and operators on the objects; it is defined similarly to the variable type of [71] and the serial data type of [37, 38]. Formally, the Obj type consists of: a set S of object states; a distinguished initial state s 0 S; a set R of response values; and . a set O of operations, each of the type S (S R) Furthermore, we assume an application defined function merge: SetOf(Proc S) S. This ....

A. Fekete, D. Gupta, V. Luchangco, N. Lynch, and A. Shvartsman. Eventuallyserializable data services. In 15th ACM Symposium on Principles of Distributed Computing (PODC), pages 300--309, May 1996.

No context found.

Fekete, A., Gupta, D., Luchangco, V., Lynch, N., and Shvartsman, A. Eventually-serializable data service. Theoretical Computer Science, 220(1):113-156, June 1999.

.... We present our incremental modification constructs in the context of the I O automata model [30, 32] the basics of the model are reviewed in Section 2) I O automata have been widely used in formulating formal service definitions and abstract implementations, and for reasoning about them, e.g. [6, 9, 11, 12, 14, 15, 21, 24, 28, 31]) An important feature of the I O automaton formalism is its strong support of composition. For example, Hickey et al. 24] used the compositional approach for modeling and verification of certain modules in Ensemble [19] a large scale, modularly structured, group communication system. ....

A. Fekete, D. Gupta, V. Luchangco, N. Lynch, and A. Shvartsman. Eventually-serializable data services. Theoretical Computer Science, special issue on Distributed Algorithms, 220, 1999.

No context found.

Fekete, A., Gupta, D., Luchangco, V., Lynch, N., and Shvartsman, A. Eventually-serializable data service. Theoretical Computer Science, 220(1):113--156, June 1999.

.... most of its proof methods appears in Chapter 8 of [50] The model and methods were originally developed for reasoning about theoretical distributed algorithms, but, in the past few years, they have been applied increasingly to practical system services such as distributed shared memory services [25, 24, 23], group communication services [26, 27, 16, 19, 41] and standard communication services like TCP [72, 71, 73] For these services, the theory has contributed system descriptions and proofs, and also important structural information such as interface specifications, invariants, and simulation ....

....Goldman s more recent work on the Programmer s Playground [33] also includes a language with formal semantics in terms of I O automata. Cheiner and Shvartsman [15] generated code for a specific I O automaton based distributed algorithm the Eventually Serializable Data Service of Luchangco et al. [24], and also gave some suggestions for more general distributed code generation strategies. Their approach was not based on a formal programming language, and did not involve theorem proving. Their principal code generation strategy requires non local synchronization for some transitions, although ....

Alan Fekete, David Gupta, Victor Luchangco, Nancy Lynch, and Alex Shvartsman. Eventuallyserializable data services. In Proceedings of the Fifteenth Annual ACM Symposium on Principles of Distributed Computing, pages 300--309, Philadelphia, PA, May 1996.

No context found.

A. Fekete, D. Gupta, V. Luchango, N. Lynch, and A. Shvartsman. EventuallySerializable Data Services. PODC 1996, pp. 300-310

No context found.

A. Fekete, D. Gupta, V. Luchangco, N. Lynch, and A. Shvartsman. Eventually-serializable data services. Theoretical Computer Science, 220(Special issue on Distributed Algorithms):113--156, 1999.

No context found.

A. Fekete, D. Gupta, V. Luchangco, N. Lynch, and A. Shvartsman. Eventually-serializable data services. Theoretical Computer Science, 220(Special issue on Distributed Algorithms):113-- 156, 1999.

No context found.

A Fekete, D Gupta, V Luchangco, N Lynch, and A Shvartsman. Eventuallyserializable data services. In PODC96 Proceedings of the Fifteenth Annual ACM Symposium on Principles of Distributed Computing, pages 300--309, 1996.

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