Distributed Shared Memory (DSM) is a collection of nodes or clusters, each with its own memory connected by an interconnected network. The key issue in DSM is keeping the memory pages consistent. It refers to the degree of consistency that has to be maintained for the shared memory data. Maintaining perfect consistency is especially painful when the difference between the latency and/or throughput of memory accesses on the one hand, and the network connecting the machines on which these copies reside on the other, is big. The solution might be to accept less than perfect consistency as the price for better performance. This paper reviews various memory consistency models which are used in different DSM systems.
Authors: Sindhu Singh
Publish Year: 2010
