Second Workshop on Real, Large Distributed SystemsPreliminary Abstract
Pp. 4954 of the Proceedings
Bridging Local and Wide Area Networks for Overlay Distributed File Systems
Mike Closson and Paul Lu, University of Alberta
Abstract
In metacomputing and grid computing, a computational
job may execute on a node that is geographically
far away from its data files. In such a situation, some
of the issues to be resolved are: First, how can the job
access its data? Second, how can the high latency and
low bandwidth bottlenecks of typicalwide-area networks
(WANs) be tolerated? Third, how can the deployment of
distributed file systems be made easier?
The Trellis Network File System (Trellis NFS) uses a
simple, global namespace to provide basic remote data
access. Data from any node accessible by Secure Copy
can be opened like a file. Aggressive caching strategies
for file data and metadata can greatly improve performance
across WANs. And, by using a bridging strategy
between the well-knownNetwork File System(NFS)
and wide-area protocols, the deployment is greatly simplified.
As part of the Third Canadian Internetworked Scientific
Supercomputer (CISS-3) experiment, Trellis NFS
was used as a distributed file system between highperformance
computing (HPC) sites across Canada.
CISS-3 ramped up over several months, ran in production
mode for over 48 hours, and at its peak, had over
4,000 jobs running concurrently. Typically, there were
about 180 concurrent jobs using Trellis NFS. We discuss
the functionality, scalability, and benchmarked performance
of Trellis NFS. Our hands-on experience with
CISS and Trellis NFS has reinforced our design philosophy
of layering, overlaying, and bridging systems to provide
new functionality.
- View the full text of this paper in HTML and PDF.
Until December 2006, you will need your USENIX membership identification in order to access the full papers. The Proceedings are published as a collective work, © 2005 by the USENIX Association. All Rights Reserved. Rights to individual papers remain with the author or the author's employer. Permission is granted for the noncommercial reproduction of the complete work for educational or research purposes. USENIX acknowledges all trademarks within this paper.
- If you need the latest Adobe Acrobat Reader, you can download it from Adobe's site.
|