Cost/performance tradeoffs in network interconnects for clusters of commodity PCs


Date

2003

Publication Type

Report

ETH Bibliography

yes

Citations

Altmetric

Data

Abstract

The definition of a commodity component is quite obvious when it comes to the PC as a basic compute engine and building block for clusters of PCs. Looking at the options for a more or less per formant interconnect between those compute nodes it is much less obvious which interconnect still qualifies as commodity and which not. We are trying to answer this question based on an in-depth analysis of a few common more or less expensive in terconnects on the market. Our measurements and observations are based on the experience of architecting, procuring and in stalling Xibalba, a 128 node - 192 processor versatile cluster for a variety of research applications in the CS department of ETH Zurich. We define our unique way to measure the performance of an interconnect and use our performance characterization to find the best cost performance point for networks in PC clusters. Since our work is tied to the purchase of a machine at fair mar ket value we can also reliably comment on cost performance of the four types of interconnects we considered. We analyze the reason for performance and non-performance for different Fast Ethernet architectures with a set of micro-benchmarks and con clude our study with performance numbers of some applications. Thus, the reader gets an idea about the impact of the intercon nect on the overall application performance in commodity PC clusters.

Publication status

published

External links

Editor

Book title

Volume

391

Pages / Article No.

Publisher

ETH Zurich, Department of Computer Science

Event

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Clusters of commodity PC; Ethernet; Myrinet; Switch performance; Application performance; Full bisection bandwidth; All-to-all communication

Organisational unit

02150 - Dep. Informatik / Dep. of Computer Science

Notes

Funding

Related publications and datasets