Mummert, L., Wing, J., and Satyanarayanan, M. Using belief to reason about cache coherence. Tech. Rep. CMU-CS-94-151, School of Computer Science, Carnegie Mellon University, May 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... [73, 8, 108] The total quantity of data transferred in an average open close session is used in publications on cache characteristics in addition to simulations of caches [98, 73] A related measure, the number of open close sessions transferring a given amount of data, has also found use [58, 68, 24]. Open close session data can also reveal such information as the frequency of use of files, the characteristics of file access (read only vs. write only vs. read write accesses) and the amount of data read from or written to files during the course of an open close session. The running of a ....

Mummert, L., Wing, J., and Satyanarayanan, M. Using belief to reason about cache coherence. Tech. Rep. CMU-CS-94-151, School of Computer Science, Carnegie Mellon University, May 1994.

....In [22] model checking is used over an abstraction of individual runs of a protocol. This abstraction reduces considerably the state space and makes model checking cheaper. However designing such an abstraction needs considerable human involvement. In fact, this paper is based on a formalization [17] made for manual proof of the same properties. Another approach is to use a theorem prover to assist manual demonstration. The work presented in [18] is based on proving properties over a protocol abstraction with a theorem prover. This abstraction is linked to the implementation through an ....

L. B. Mummert, J. M. Wing, and M. Satyanarayanan. Using belief to reason about cache coherence. In 13th ACM Conference on Principles of Distributed Computing, Los Angeles, CA, USA, Aug. 1994.

....by abstracting the system and checking a finite model of it. In this paper, we consider model checking cache coherence protocols for two distributed file systems, the Andrew File System (AFS) and the Coda File System. Mummert, Wing and Satyanarayanan derived abstract models of these protocols [9]. They also specified a cache coherence invariant (CC) Our goal is to check whether these models satisfy CC, using SMV. For this goal, we perform certain application specific abstractions to reduce the size of the corresponding SMV input programs. We consider four models AFS0, AFS1, AFS2 and ....

....For this goal, we perform certain application specific abstractions to reduce the size of the corresponding SMV input programs. We consider four models AFS0, AFS1, AFS2 and Coda . AFS0 is a simple model on which all the others are based. Models for AFS1 and AFS2 were defined by Mummert et al. [9]. Coda is the version of the cache coherence protocol for Coda developed by Mummert and Satyanarayanan [8] Section 2 presents these models and CC. Section 3 describes the SMV input language by giving a simple example. Section 4 presents details of the verification of the four models. Section 5 ....

[Article contains additional citation context not shown here]

Mummert, L., Wing, J., and Satyanarayanan, M. Using belief to reason about cache coherence. In Proceedings of the Symposium on Principles of Distributed Computing (Aug. 1994). Also CMU-CS-94-151, May 1994.

....strategy as an obligation that the server mantains towards its clients. That obligation is understood by both parts and serves as a contract. The mere understanding of the contract gives clients a lot of information on how to handle their caches. This notion is similar to the agreement set in [12] which determines for each data object whether the server and client copies are equal. In [12] the authors use logic of authentication [5] and the notion of belief to reason about the cache coherence provided by some protocols. It is possible to use such tools to reason about the protocols ....

....understood by both parts and serves as a contract. The mere understanding of the contract gives clients a lot of information on how to handle their caches. This notion is similar to the agreement set in [12] which determines for each data object whether the server and client copies are equal. In [12], the authors use logic of authentication [5] and the notion of belief to reason about the cache coherence provided by some protocols. It is possible to use such tools to reason about the protocols presented in this paper as well. 3 The replication of the database between many servers is ....

L. Mummert, J.M. Wing, and M. Satyanarayanan. Using Belief to Reason About Cache Coherence. In Proceedings of the ACM Sigact-Sigops Symposium on the Principles of Distributed Computing, August 1994.

....terms of a well known theory, it pays to reuse that work and not invent from scratch. If it is not standard then an informal statement of the correctness condition should drive the formalization of the system model and expression of the correctness condition. For example, in work by Mummert et al. [13], the authors started with this informal statement of cache coherence for a distributed file system: If a client believes that a cached file is valid then the server that is the authority on that file had better believe that the file is valid. They developed a system model (a state machine model) ....

L. Mummert, J.M. Wing, and M. Satyanarayanan. Using belief to reason about cache coherence. In Proceedings of the Symposium on Principles of Distributed Computing, pages 71--80, August 1994. Also CMU-CS-94-151, May 1994.

....file and its volume, the run ends when the file is discarded or rendered suspect. A failure or an invalidation for the file is sufficient to end the run. An invalidation for the volume ends the run only if there is no callback on the file. 2 A complete analysis of all three protocols appears in [18]. Without loss of generality, we can analyze this cache coherence protocol by considering one client, C , one server, S , one file, f , and one volume, V , with version number v. Implicitly, the system includes at least one other client to represent remote updates. We can capture the system ....

Lily B. Mummert, Jeannette M. Wing, and M. Satyanarayanan. Using Belief to Reason About Cache Coherence. Carnegie Mellon University, School of Computer Science Technical Report, in preparation.

....to achieve than on a multiprocessor because of the presence of failures and transmission delay. Thus, since global knowledge is impossible to achieve in a distributed system [18] we settle for pairwise knowledge between clients and servers. Our notion of belief captures this pairwise knowledge [25]. An invariant property to prove of all cache coherence protocols is that if a client believes that a cached file is valid, then the server that is the authority on the file believes the client s copy is valid. More formally, CC: 8 C : client : 8 S : server : 8 f : f ile : C believes valid(f C ) ....

....is that if a client believes that a cached file is valid, then the server that is the authority on the file believes the client s copy is valid. More formally, CC: 8 C : client : 8 S : server : 8 f : f ile : C believes valid(f C ) S believes valid (f C ) where fC stands for the copy of f at C [25]. A validation based technique, AFS 1, was used in the Andrew File System from 1984 1985 [27] for performance reasons, an invalidation based technique, AFS 2, replaced AFS 1 and has been in use in Andrew since 1985 [20] In 1993 Mummert, as part of her Ph.D. thesis work, started implementing a ....

[Article contains additional citation context not shown here]

Mummert, L., Wing, J., and Satyanarayanan, M. Using belief to reason about cache coherence. In Proceedings of the Symposium on Principles of Distributed Computing (Aug. 1994). Also CMU-CS-94151, May 1994.

....terms of a well known theory, it pays to reuse that work and not invent from scratch. If it is not standard then an informal statement of the correctness condition should drive the formalization of the system model and expression of the correctness condition. For example, in work by Mummert et al. [16], the authors started with this informal statement of cache coherence for a distributed file system: If a client believes that a cached file is valid then the server that is the authority on that file had better believe that the file is valid. They developed a system model (a state machine model) ....

L. Mummert, J.M. Wing, and M. Satyanarayanan. Using belief to reason about cache coherence. In Proceedings of the Symposium on Principles of Distributed Computing, pages 71--80, August 1994. Also CMU-CS-94-151, May 1994.

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