Volume 7, Issue 1 -
Spring/Summer 1999

Volume 6, Issue 3
Fall 1998

Volume 6, Issue 2
Spring/Summer 1998

Volume 6, Issue 1
Winter 1998

Volume 5, Issue 4
Fall 1997

Volume 5, Issue 3
Summer 1997

Volume 5, Issue 2
Spring 1997

Volume 5, Issue 1
Winter 1997

Volume 4, Issue 4
Fall 1996

Volume 4, Issue 3
Summer 1996

Volume 4, Issue 2
Spring 1996

Volume 4, Issue 1
Winter 1996

Volume 3, Issue 4
Fall 1995

Volume 3, Issue 3
Summer 1995

Volume 3, Issue 2
Spring 1995

Volume 3, Issue 1
January 1995

Volume 2, Issue 4
October 1994

Volume 2, Issue 3
July 1994

Volume 2, Issue 2
April 1994

Volume 2, Issue 1
January 1994

Volume 1, Issue 4
October 1993

Volume 1, Issue 3
July 1993

Volume 1, Issue 2
April 1993

Volume 1, Issue 1
January 1993


Paul Messina, Mani Chandy, Carl Kesselman, Tony Lenard, Dan Meiron, Vince McKoy, Jim Pool, Thomas Prince, Jim Bower, Aaron Kuppermann, Caltech; Dan Reed, Andrew Chien, University of Illinois, Urbana-Champaign.

One of the next technical challenges in high-performance computing is to ensure that input/output (I/O) data rates match increasing rates of computation. With the transfer of data often dominating execution, I/O is rapidly emerging as one of the biggest performance limitations for running scientific applications on parallel computers. This group is developing strategies to improve performance of I/O intensive applications.

Their specific approach involves characterizing I/O for a set of parallel applications, abstracting and defining I/O templates and application-level methodologies, and implementing and testing I/O tools on large-scale computations. To develop appropriate designs for scalable I/O, the group is studying the I/O dynamics of a range of current applications on massively parallel systems and how parallel systems respond to application I/O requests. These applications include computational chemistry, computational fluid dynamics, computational biology, computational astronomy, and materials processing. Illinois' Pablo performance analysis environment is the basis of this I/O characterization study. Using the analysis of current I/O patterns and system responses, together with extrapolations of I/O needs from application scientists, the group's goal is to develop a set of I/O templates and libraries that support common I/O patterns.

Much of this work has benefited from the interaction of group members with other CRPC researchers involved with scalable I/O projects. At Illinois, Reed is also involved with an ARPA I/O characterization project and a project with NASA Goddard on analysis of satellite image processing codes. In addition, CRPC researchers at Argonne, Rice, Illinois, Maryland, and Caltech are founding partners in the Scalable I/O Initiative (SIO), a consortium of academic, government, and industry research and development groups with ongoing research projects on key aspects of the I/O problem. The goal of the Scalable I/O Initiative is to ensure the balance between compute power and I/O in scalable parallel processors, in particular, future teraFLOPS systems. This goal will be accomplished through a coordinated attack using full- scale applications on full-scale testbeds to characterize I/O and to motivate and evaluate changes in languages, compilers, operating systems, and file systems.

Table of Contents