Distributed Shared Memory: To Relax or Not to Relax?

Iosevich, Vadim; Schuster, Assaf

doi:10.1007/978-3-540-27866-5_25

Cited by 1 publication

(1 citation statement)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[9] focuses on reducing communication required to maintain consistency among distributed memories. [28,57,68] and others try to relax consistency using basic memory models previously described. Since they inherently aim to provide a consistent DSM, relaxation of consistency is not explored in these works.…”

Section: Related Workmentioning

confidence: 99%

Aspire

Vora

Koduru

Gupta

2014

Proceedings of the 2014 ACM International Conference on Object Oriented Programming Systems Languages &Amp; Applications

View full text Add to dashboard Cite

Many vertex-centric graph algorithms can be expressed using asynchronous parallelism by relaxing certain read-afterwrite data dependences and allowing threads to compute vertex values using stale (i.e., not the most recent) values of their neighboring vertices. We observe that on distributed shared memory systems, by converting synchronous algorithms into their asynchronous counterparts, algorithms can be made tolerant to high inter-node communication latency. However, high inter-node communication latency can lead to excessive use of stale values causing an increase in the number of iterations required by the algorithms to converge. Although by using bounded staleness we can restrict the slowdown in the rate of convergence, this also restricts the ability to tolerate communication latency. In this paper we design a relaxed memory consistency model and consistency protocol that simultaneously tolerate communication latency and minimize the use of stale values. This is achieved via a coordinated use of best effort refresh policy and bounded staleness. We demonstrate that for a range of asynchronous graph algorithms and PDE solvers, on an average, our approach outperforms algorithms based upon: prior relaxed memory models that allow stale values by at least 2.27x; and Bulk Synchronous Parallel (BSP) model by 4.2x. We also show that our approach frequently outperforms GraphLab, a popular distributed graph processing framework.

show abstract

Section: Related Workmentioning

confidence: 99%