Computer Science

Colloquium Details

Smart File Objects

Abstract

In this talk we introduce a new paradigm for remote file access called Smart File Objects (SFOs). SFOs are designed to address the network bottleneck inherent in wide-area and wireless networks. In particular, remote file access can be inefficient due to high latency and low bandwidth. It can also be unpredictable due to fluctuating latency and bandwidth, and unreliable due to network outages. The SFO is application-specific middleware inserted on the "edge of the network" between the application and remote file. The SFO performs adaptive caching and prefetching on behalf of the application to mitigate the impact of the network. It can also offer additional advantages including high-level APIs, optimized parallel I/O, and non-blocking I/O.

The SFO is particularly well-suited to metacomputing in which high performance applications running in one site access data from another site. We describe the design of SFOs for several applications including a distributed gene sequence comparison code running across the Internet and the vBNS and an MPEG player accessing movies across the Internet. The results indicate that the SFO can provide real performance gains for a variety of applications and file types.

Biography

Jon B. Weissman received the B.S. degree from Carnegie-Mellon University in 1984, and the M.S. and Ph.D. degrees from the University of Virginia in 1989 and 1995 respectively, all in Computer Science. He has been an Assistant Professor of Computer Science at the University of Texas at San Antonio since 1995. He was an active member of the Mentat and Legion research groups while at the University of Virginia. His current research interests are in distributed systems, high performance computing, and metacomputing.