Relaxed consistency and coherence granularity in DSM systems

Abstract

During the past few years, two main approaches have been taken to improve the performance of software shared memory implementations: relaxing consistency models and providing fine-grained access control. Their performance tradeoffs, however, we not well understood. This paper studies these tradeoffs on a platform that provides access control in hardware but runs coherence protocols in software, We compare the performance of three protocols across four coherence granularities, using 12 applications on a 16-node cluster of workstations. Our results show that no single combination of protocol and granularity performs best for all the applications. The combination of a sequentially consistent (SC) protocol and fine granularity works well with 7 of the 12 applications. The combination of a multiple-writer, home-based lazy release consistency (HLRC) protocol and page granularity works well with 8 out of the 12 applications. For applications that suffer performance losses in moving to coarser granularity under sequential consistency, the performance can usually be regained quite effectively using relaxed protocols, particularly HLRC. We also find that the HLRC protocol performs substantially better than a single-writer lazy release consistent (SW-LRC) protocol at coase granularity for many irregular applications. For our applications and platform, when we use the original versions of the applications ported directly from hardware-coherent shared memory, we find that the SC protocol with 256-byte granularity performs best on average. However, when the best versions of the applications are compared, the balance shifts in favor of HLRC at page granularity.

Software Distributed Shared Memory: a VIA-based implementation and comparison of sequential consistency with home-based lazy release consistency

“…Another study was conducted by Zhou et al [23] on non-commodity hardware that supports access control at custom granularity. This allows a uniform access control mechanism to be used for both fine-grain and coarse-grain protocols.…”

“…This is because outstanding differences in implementations, testbed environments, and benchmark suites make it difficult to compare results. We are aware of three studies that compare sequentially consistent coherence protocols to relaxed consistency protocols on the same platform [22][23][24]. The section Comparing SC to Relaxed Consistency Models details these studies and outlines their differences from our work.…”

Impact of Cache Coherence Models on Performance of OpenMP Applications

“…Relaxed consistency models [12] formalize the use of inconsistent memory hence giving the programmer a consistent memory abstraction. They are generally combined with synchronization primitives, like locks and barriers, which are contained in shared memory codes.…”

Distributed Shared Memory: To Relax or Not to Relax?

“…Additional research was conducted by Zhou et al [20] on a noncommodity hardware system that supports access control at custom granularity. This allows the use of a uniform access control mechanism for both fine-grain and coarsegrain protocols.…”

“…The combination of a HLRC protocol and page granularity works well with 8 of the 12 applications. It should be noted that the testbed environment used in [20] consisted of 66MHz processors, while our processors are an order of magnitude faster. In contrast to this study, we use standard hardware to implement DSM protocols.…”