The Scalable I/O Initiative, a collaboration of agencies, vendors, and institutions that includes CRPC researchers across the country, was formed in 1994 to overcome the biggest obstacle to effective use of teraflops-scale computer systems by scientists and engineers--getting data into, out of, and around such systems fast enough to avoid severe bottlenecks. The initiative is determining applications requirements and using them to guide the development of system support services, file storage facilities, high-performance networking software, programming language features, and compiler techniques.
As part of an ongoing program to inform and solicit input from the HPCC community, Initiative members held the workshop "Research Issues in Scalable I/O" at SUPERCOMPUTING '95. Participants examined research issues, evaluated progress, and described open problems of the initiative. CRPC-related presenters included Paul Messina and Jim Pool of the California Institute of Technology, Chuck Koelbel of Rice University, and Dan Reed of the University of Illinois at Urbana-Champaign.
Messina opened the workshop with an overview of the Initiative, after which Pool described the collection of I/O intensive applications under investigation. Reed reported on the initial experience in instrumenting, measuring, and analyzing these applications. Kai Li of Princeton University described the plans for operating systems, file systems, and networking. In particular, he outlined the progress toward the Scalable I/O Initiative's Parallel I/O Application Program Interface, initially a low-level interface to the operating system for constructing such components as file systems and runtime libraries. Rusty Lusk and Rajeev Thakur of Argonne National Laboratory presented the scalable I/O methodology for characterizing I/O benchmarks and how it drives other project developments, using as an example an astrophysics computation. Koelbel illustrated the plans for compilers and languages using the example of out-of-core computations.
The workshop served as a starting point for a project technical meeting held in January 1996. This meeting extended the workshop discussion to explore additional details, including developing the schedule for evolving and reviewing the low-level version of the Scalable I/O Initiative's Parallel I/O Application Program Interface.
For more information about the Scalable I/O Initiative, see http://www.cacr.caltech.edu/SIO/ or http://www.mcs.anl.gov/scalable/scalable.html .