V. V. Murty and V. K. Garg. Characterization of message ordering specifications and protocols. In Proceedings of the 17th International Conference on Distributed Computing Systems, May 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....possible sequences in which the channel may deliver the multicast messages. Many existing frameworks provide support for FIFO, causal and total ordering semantics [BvR94, MMSA 96, ADKM92, KS98] Application semantics has been exploited in group communication frameworks to improve performance [MG97, Sch88, JSM97, LLS90] For example, replicas of a database can be maintained consistent by delivering the operations to all replicas in the same order. However, in some cases, the application entities may be able to accept operations in a sequence satisfying a weaker ordering requirement and ....

....may be able to accept operations in a sequence satisfying a weaker ordering requirement and still maintain consistency (some operations may be commutative and can be delivered in either order) Such weakening of ordering requirements have shown to result in performance improvements. JSM97, MG97, Sch88] have proposed notions such as conflict predicates and forbidden predicates to communicate such application information to the GCS. In this paper, we study the further use of application semantics to configure a GCS. In general, a GCS does not make any assumption regarding the ....

V. Murty and V. Garg. Characterization of message ordering specifications and protocols. In Proc. IEEE International Conference on Distributed Computing Systems

....computation. Off line control is easier to achieve than on line control because of the pre knowledge of future events. One method to control a computation is to re order messages. This form of control (which may also be considered a testing methodology) has been studied in both its on line [12] and off line [8] variants. We consider a less intrusive form of control which is only capable of delaying events (and not reordering them) This has been termed the predicate control problem and has (a) Good Computation a b c d (b) Bad Computation CS c d a b CS CS (c) New model c d a b ....

V. V. Murty and V. K. Garg. Characterization of message ordering specifications and protocols. In Proceedings of the 17th International Conference on Distributed Computing Systems, May 1997.

....at certain points are send and delivery, then the protocol must enable at least one of the events. This restriction follows from our assumption of online control. Since the future is not known, the controller must always enable one of the current events for progress. The following result in [MG97b] shows the limitations of the controllers based on above model. We call a computation synchronously ordered if all messages can be drawn vertically in its process time diagram. We first argue that any computation that is synchronously ordered must be allowed by the protocol. Since the computation ....

....We now repeat this sequence with the second message. Thus, due to liveness condition and the fact that all messages can be sorted, we get that the synchronous computation is possible under the protocol. One way for the programmer to specify desired message ordering is by using forbidden predicates [MG97b]. A message ordering is acceptable only if it does not satisfy the given forbidden predicate. A forbidden predicate is a conjunction of causality relationships between sends and receives of messages. For example, the following forbidden predicate specifies violation of causal ordering: 9x; y : ....

[Article contains additional citation context not shown here]

V. V. Murty and V. K. Garg. Characterization of message ordering specifications and protocols. In Proc. of the International Conference on Distributed Computing Systems, page to appear. IEEE, May 1997.

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