Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Towards virtual shared memory for non-cache-coherent multicore systems
Ramesh B., Ribbens C., Varadarajan S.  IPDPSW 2013 (Proceedings of the 2013 IEEE 27th International Symposium on Parallel and Distributed Processing Workshops, May 20-24, 2013)1186-1193.2013.Type:Proceedings
Date Reviewed: Dec 4 2013

Samhita, an implementation of distributed shared memory (DSM), is discussed in this paper. DSM enables a programmer to view a collection of compute nodes, each with its own physical memory, as a single shared memory, greatly simplifying programming. Efficient implementation of DSM was an active research area some time ago, but it failed to find widespread acceptance due to the long latency of the interconnection between nodes. However, DSM is making a comeback due to the recent availability of high-speed commodity interconnects with low latency.

The authors begin by making a case for the adoption of DSM for connections between heterogeneous computing elements of a single computer, such as graphics processing units (GPUs) and central processing units (CPUs). The rest of the paper describes the performance analysis of the Samhita DSM, which was built by the authors and described in a different paper.

The paper provides a welcome contribution in its analysis with microbenchmarks of the Samhita DSM, which might be of interest to practitioners in the area. The authors show that Samhita scales well and has low overhead using a small set of benchmarks and a two-node cluster.

The organization and content of the paper is a little underwhelming from a general perspective. DSM concepts have been well studied, and the paper really shows nothing new with Samhita. There is also a lack of performance comparisons with other DSMs. Furthermore, the paper starts with the intention to target a heterogeneous system, but mentions nothing else about it. Instead, the authors conduct their performance evaluation using a cluster with two physical machines. Some plans are laid out for future work for heterogeneous systems, but these are not very useful for readers who might have been interested in a DSM for heterogeneous systems.

Reviewer:  Amitabha Roy Review #: CR141780 (1402-0131)
Bookmark and Share
 
Shared Memory (B.3.2 ... )
 
 
Buses (C.2.5 ... )
 
 
Distributed Architectures (C.1.4 ... )
 
 
High-Speed (C.2.5 ... )
 
 
Network Communications (C.2.1 ... )
 
 
Data Communications Devices (B.4.1 )
 
Would you recommend this review?
yes
no
Other reviews under "Shared Memory": Date
Scalable shared-memory multiprocessor architectures
Thakkar S., Dubois M., Laundrie A., Sohi G. (ed) Computer 23(6): 71-73, 1990. Type: Article
Jul 1 1991
Memory coherence in shared virtual memory systems
Li K., Hudak P. ACM Transactions on Computer Systems 7(4): 321-359, 1989. Type: Article
Oct 1 1990
Multigrain shared memory
Yeung D., Kubiatowicz J., Agarwal A. ACM Transactions on Computer Systems 18(2): 154-196, 2000. Type: Article
May 1 2001
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy