Abstract:
This paper describes the comparison between homeless and home-based Lazy Release Consistency (LRC) protocols which are used to implement Distributed Shared Memory (DSM) in cluster computing. We present a performance evaluation of parallel applications running on homeless and home-based LRC protocols. We compared the performance between Tread-Marks, which uses homeless LRC protocol, and our home-based DSM system. We found that the home-based DSM system has shown better scalability than TreadMarks in parallel applications we tested. This poor scalability in the homeless protocol is caused by a hot spot and garbage collection, but we have shown that these factors do not affect the scalability of the home-based protocol.
