Parallel Profile

Carl Kesselman

Project Leader, University of Southern California Information Sciences Institute; and Visiting Associate, Computer Science Department, California Institute of Technology

Carl Kesselman's research focuses on developing new methods, tools, and programming environments for large-scale, high performance computer systems. His research has resulted in new techniques for performance measurement and visualization of parallel programs, innovative new programming languages, such as Compositional C++, library structures for building reusable parallel program components, and new techniques for supporting multithreaded, dynamic parallel computations, such as Nexus.

Kesselman received his B.S. in electrical engineering and B.A. in computer science from the University of Buffalo (1982), his M.S. in electrical engineering at the University of Southern California (1984), and his Ph.D. in computer science from the University of California at Los Angeles (1991). He began his career in 1982 as a member of the technical staff in the Computer Science Laboratory of the Aerospace Corporation, in El Segundo, California, where he initiated an ongoing research program in high performance computing. While at Aerospace, Kesselman focused on parallel logic programming systems, developing tools for performance measurement and performance visualization of parallel logic programs.

In 1991, Kesselman joined the California Institute of Technology, first as a senior research fellow in computer science (1991 to 1993) and then as a member of the Beckman Institute (1993 to 1996). Kesselman remains on the faculty of Caltech as a visiting associate in computer science. As part of the CRPC, Kesselman, in collaboration with K. Mani Chandy, designed and implemented the Compositional C++ parallel programming language. This language is currently being used for research projects in parallel computation at a number of institutions all over the world. "The goal of CC++ is to allow the integration of many different parallel programming styles into a single program," says Kesselman. "Much of the focus in parallel programming language such as HPF has been to support loosely synchronous behavior and a basically homogeneous structure. CC++ was designed to support more general types of concurrency. Because of this flexibility, CC++ is an ideal language for building high- performance distributed computing or metacomputing applications, which are increasing in importance."

In 1996, Kesselman joined the Information Sciences Institute of the University of Southern California (ISI) to lead a research group in high-performance parallel and distributed systems. Kesselman's current research focuses on understanding how to construct high-performance computations that couple high-performance computers with state-of-the- art high-speed networks.

As part of this research, Kessleman co-leads the Globus project, which is investigating core technologies for resource discovery, scheduling, configuration, security, data access, and execution in high performance networked environments (see "Research Focus," page 8). He also leads the Qualis project, which is investigating how to incorporate network quality of service requirements into parallel applications. "I believe networking is the future of high-performance computation," he says. "It enables computations to be viewed from a systems perspective in which specialized components, such as data-archives, massively parallel computers, visualization engines, and data sources can be integrated into a single computation environment. The problem lies in understanding the software abstractions and building the infrastructure that will allow you to achieve distribution without sacrificing too much performance."

Kesselman has also spent time as a visiting scientist at the Swedish Institute of Computer Science, the Mathematics and Computer Science Division at Argonne National Laboratory, and ICASE, at NASA Langley.

Kesselman has authored or co-authored more than 25 journal articles, book chapters, and papers. He is a member of the technical steering committee of the CRPC and has served on numerous technical program committees, including the International Workshop on Parallel C++ (1996), Frontiers of Parallel Computation (1994 and 1996), Supercomputing '94, and Supercomputing '97.

Table of Contents