A Distributed Shared-Memory System on a Workstation Cluster Using Fast Serial Links

Authors:
Hironori Nakajo;Akihiro Ichikawa;Yukio Kaneda
Affiliations:
Division of Computer Information and Communication Sciences, Tokyo University of Agriculture and Technology;Department of Computer and Systems Engineering, Kobe University;Department of Computer and Systems Engineering, Kobe University
Venue:
International Journal of Parallel Programming - Special issue on international symposium on high performance computing 1997, part I
Year:
2000

Citing 9
Cited 0

Memory coherence in shared virtual memory systems

ACM Transactions on Computer Systems (TOCS)
PVM: a framework for parallel distributed computing

Concurrency: Practice and Experience
Implementation and performance of Munin

SOSP '91 Proceedings of the thirteenth ACM symposium on Operating systems principles
Lazy release consistency for software distributed shared memory

ISCA '92 Proceedings of the 19th annual international symposium on Computer architecture
The SPLASH-2 programs: characterization and methodological considerations

ISCA '95 Proceedings of the 22nd annual international symposium on Computer architecture
An I/O network architecture of the distributed shared-memory massively parallel computer JUMP-1

ICS '97 Proceedings of the 11th international conference on Supercomputing
Scalable Shared-Memory Multiprocessing

Scalable Shared-Memory Multiprocessing
Myrinet: A Gigabit-per-Second Local Area Network

IEEE Micro
Tempest: a substrate for portable parallel programs

COMPCON '95 Proceedings of the 40th IEEE Computer Society International Conference

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, a fast serial link, Serial Transparent Asynchronous First-in First-out Link (STAFF-Link), is introduced. Using such links, we construct a parallel processing system based on a workstation cluster. The workstation cluster implements a distributed sharedmemory mechanism for inter-process communication using a software controlled cache using a STAFF-Link router board. The board has a chained multicast capability with which we have implemented efficient invalidation protocol based on Eager Release Consistency (ERC) model in the DSM system. Performance results on several application programs from the SPLASH2 benchmark suites have been measured.