Distributed Shared Memory based on Group Large Causality
The implementation of Distributed Shared Memory (DSM) must be simple and efficient to become a general purpose tool. In general, known DSM systems use page-oriented virtual memory systems to simulate the shared address space. The usual implementation uses vector clocks or history tuples associated to pages, being very expensive and impossible to scale to large systems.
In this paper, we show a novel approach to this implementation problem, proposing an algorithm based on a relaxed causal ordering of multicasts and messages (called Group Large Causality) to provide Coherent Causal Consistency DSM in large-scale networks.
KeywordsDistributed Shared Memory Causality Replica Coherency
Unable to display preview. Download preview PDF.
- 1.M. Ahamad, G. Neiger, J. Burns, P. Kohli, P. Hutto, “Causal Memory: Definitions, Implementation and Programming,” Tech. Report GIT-CC-93/55, College of Computing, Georgia Institute of Technology, 1993.Google Scholar
- 2.K. P. Birman, T. A. Joseph, “Exploiting Replication in Distributed Systems,” in Distributed Systems, edited by S. Mullender, ACM Press, N.Y., 1989.Google Scholar
- 3.A. Campos, J. Navarro, “Coherent Causal Consistency in Distributed Shared Memory,” Proc. XV International Conference of the Chilean Computer Science Society, Arica, Chile, October 1995.Google Scholar
- 4.K. Garachorloo, K. Lenoski et al., “Memory Consistency and Event Ordering in Scalable Shared Memory Multiprocessors,” Proc. 17th Annual International Symposium on Computer Architecture, May 1990.Google Scholar
- 5.J. Goodman, “Cache Consistency and Sequential Consistency,” Tech. Report 61, IEEE Scalable Coherent Interface Working Group, 1989.Google Scholar
- 6.P. Hutto, M. Ahamad, “Slow Memory: Weakening Consistency to Enhance Concurrency in Distributed Shared Memories,” Proc. IEEE 10th International Conference on Distributed Computing Systems, 1990.Google Scholar
- 7.R. John, M. Ahamad, “Evaluation of Causal Distributed Shared Memory for Data-Race-Free Programs”, Tech. Report GIT-CC-94/34, College of Computing, Georgia Institute of Technology, 1993.Google Scholar
- 10.J. M. Piquer, “Indirect Reference Counting: A Distributed GC,” LNCS 505, PARLE '91 Proceedings Vol I, pp. 150–165, Springer Verlag, Eindhoven, The Netherlands, June 1991.Google Scholar
- 11.J. M. Piquer, “Large Causality: Ordering Broadcasts and Messages,” position paper, 5th ACM SIGOPS Workshop on Models and Paradigms for Distributed Systems Structuring, Mt Saint-Michel, France, September 1992.Google Scholar
- 12.J. M. Piquer, C. Queinnec, “TransPive: A Distributed Lisp System,” Lettre du Transputer, Laboratoire d'Informatique de Besançon, N. 16, pp. 55–68, December 1992.Google Scholar
- 13.J. M. Piquer, “A Reimplementation of TransPive: Lessons from the Experience,” Parallel Symbolic Languages and Systems (PSLS'95), Beaune, France, October 1995.Google Scholar