Evaluating the Effect of Contention on Page-Based Software Shared Memory Systems

In this paper, we explore experimentally the effects of contention on the performance of page-based software distributed shared memory systems using lazy release consistency. We use two existing protocols, Princeton's home-based protocol and the TreadMarks protocol, and a third novel protocol, Adaptive Striping. Our results show that programs exhibiting good scalability experience minor variations in memory latency (for example, only a 4.5% increase as we quadruple the number of nodes), do not suffer from contention, and have a good balancing of communication. In contrast, programs that scaled poorly suffered from large memory latency increases (up to 480% higher) due to contention and communication imbalances. For most of our programs, all three protocols were equally affected by latency increases and achieved similar performance. Where they differ significantly, the protocol load imbalance (caused by the read accesses to pages that have multiple readers following one or more writers)..