NSF CISE Research Infrastructure Program Award Number: CDA-9703088 Parallel and Distributed Computing: Systems and Application Development Infrastructure Progress Report for Period September 1, 1997 - June 1, 1998 PIs: Lori Pollock, Guang Gao, Sandra Carberry, and Errol Lloyd Affiliated Investigators: Gagan Agrawal, Gonzalo Arce, Daniel Chester, Keith Decker, Chandra Kambhamettu, Ashfaq Khokhar, David Saunders, Adarsh Sethi, Darren Vengroff Department of Computer and Information Sciences Department of Electrical and Computer Engineering University of Delaware Newark, DE 19716 302 831-1953 fax: 302 831-8458 1 Introduction The University of Delaware, marking its 255th year as the only research university in the State of Delaware, is a state-assisted, privately governed institution with approximately 15,000 undergraduate students, 3,000 graduate students, 3,000 students enrolled in credit courses through Continuing Education, and 900 faculty members. The Department of Computer and Information Sciences (CIS) is one of 25 departments in the College of Arts and Sciences, and offers BA, BS, MS, and PhD degrees. There are 16 full-time tenure- track or tenured faculty, 2 visiting faculty, and 5 research faculty in the department. There are 320 CIS undergraduate majors and 90 CIS graduate students. CIS courses are also taken by non-major students both as requirements in their degree programs and to earn a CIS minor. The Department of Electrical and Computer Engineering (ECE) is one of 6 departments in the College of Engineering, and offers a Bachelor of Electrical Engineering, Bachelor of Computer Engineering, MS and PhD degrees. There are 5 full-time tenure-track or tenured faculty in the Computer Engineering portion of ECE, and there are 65 undergraduate Computer Engineering majors, along with 15 Computer Engineering graduate students. While faculty research in the University of Delaware CIS and ECE departments spans a wide range of areas, a large cross section of the faculty are actively pursuing research involving parallel and/or distributed systems. In particular, Delaware has a core group of faculty who specialize in compiler optimization, run- time systems, architectural and algorithm design for high performance computing systems. Several faculty work in developing software for improved services on a variety of distributed applications, utilizing the new generation of high-speed network switches. Further, a number of application projects underway at Delaware are aimed at parallel and/or distributed environments. These applications include parallel symbolic analysis, video-image processing, real-time multi-agent systems, gesture recognition, and multimodal applications. Over the past several years, the faculty research focus has been on theoretical approaches that model the key aspects of their research problems, and on simulation studies that are implemented versions of the theoretical models. These approaches, while extremely valuable in elucidating the fundamental structure of, and approaches to, a problem, cannot include all of the many and varied aspects that arise in practice. In this context, there has been an urgent need for this ongoing research to incorporate realistic and robust experimental studies. The NSF CISE Research Infrastructure Award described here is aimed at establishing a general infrastructure in the University of Delaware CIS and ECE departments for experimental research in parallel and distributed system software tools, and in parallel and distributed applications development. Prior to this grant, there were no permanent experimental resources available to CIS and ECE faculty. As a result of the developments of faster workstations and networking switches, the areas of systems, networking, and parallel applications can all share identical infrastructure on which to perform experimental research. In particular, the presence of a local, changeable, crashable computing resource of substantial size is a particu- lar benefit. This grant provides the necessary funds to construct a research infrastructure based on a cluster of high-performance, multiprocessor workstations connected via high-speed networks. Three primary goals of the grant are: