Efficient parallel processing on low-cost clusters with GAMMA active ports

Citation
G. Chiola et G. Ciaccio, Efficient parallel processing on low-cost clusters with GAMMA active ports, PARALLEL C, 26(2-3), 2000, pp. 333-354
Citations number
18
Categorie Soggetti
Computer Science & Engineering
Journal title
PARALLEL COMPUTING
ISSN journal
01678191 → ACNP
Volume
26
Issue
2-3
Year of publication
2000
Pages
333 - 354
Database
ISI
SICI code
0167-8191(200002)26:2-3<333:EPPOLC>2.0.ZU;2-Z
Abstract
The Genoa Active Message MAchine (GAMMA) is an efficient communication laye r for 100base-T clusters of Personal Computers under the Linux operating sy stem (OS), It is based on Active Ports, a communication mechanism derived f rom Active Messages. Active Ports share most of the low-level optimization opportunities with Generic Active Messages while offering a higher-level pr ogramming interface not only in the SPMD but also in the MIMD and client/se rver paradigms. In addition to point-to-point communications, multi-cast, b arrier synchronization, scatter, and gather primitives have also been devel oped based on Active Ports and exploiting shared 100base-T LAN technology i n an optimal way. GAMMA Active Ports deliver excellent communication perfor mance at the user level (latency 13 mu s, maximum throughput 12.2 MByte/s, half-power point reached with 200 byte long messages), thus enabling cost-e ffective cluster computing on 100base-T. Despite being implemented at the k ernel level in the Linux OS, performance numbers of GAMMA Active Ports are much better than many other LAN-oriented communication layers, including so called "user-level" ones (e.g. U-Net). Some code porting efforts have alre ady shown that several applications are reasonably easy to develop on top o f GAMMA and that they can actually take advantage of the efficient point-to -point as well as collective communication primitives offered by our protot ype library implementation. A porting of the MPICH higher-level interface a top GAMMA is currently under way. (C) 2000 Published by Elsevier Science B. V. All rights reserved.