A Simulation_Based Comparison between Migration and Full Replication Dsm Algorithms

Distributed Shared Memory is a memory scheme for a distributed system where a collection of workstations share a single, paged virtual address space. Needed pages are obtained from other processors over the network.In its simplest version, each page is present on exactly one machine. References to local pages are done with hardware assist at full memory speed. An attempt to reference a page on another machine stimulates the Distributed Operating System ( DOS ). The DOS sends a message to the remote machine, which finds the needed memory page and sends it to the requesting processor.All communication and synchronization can be done in "memory", without explicit communication visible to user processes. The system is easy to program (logically shared memory) and easy to build (no physically shared memory).This study is concerned with a simulation comparison between migration and full replication algorithms. The simulation program calculates the response time and the throughput against number of nodes and workload depending on a set of arguments like DSM page size (Granularity), atomic search time, timeout (Quantum), and message size.The full replication is the second level extension of the migration algorithm. The migration algorithm depends on the migration of the page from node to node without the use of the replication of the page, while the full replication algorithm depends on the page replication among all nodes depending on the central control from the sequencer to ensure that the write request is sequenced in order to keep data consistency