9625291 Shahabuddin The research plan for this project focuses on new strategies for efficient simulation of computer and telecommunication systems. For many large-scale systems, simulation is one of the few means available to analyze system performance, but because these systems must often meet critical standards, estimating the probabilities of certain "rare events", e.g. overflows from a finite buffer queue or system failures, via simulation may be time-consuming and difficult. The research has two main thrusts: first, to investigate the extension of importance sampling techniques to non-Markovian processes, and second, to investigate a relatively new technique called importance splitting. These techniques will be studied in the context of queueing and reliability problems. Collaboration with researcher(s) at a major computer systems manufacturer is proposed on performance analysis of high speed telecommunications networks." The educational plan has two specific goals (1) to develop an advanced simulation course at the M.S./Ph.D. level that would expose students to more modeling and analysis tools; and (2) to expand the involvement of talented undergraduate students in research activities. The investigator will enhance the exposure of students to industrial problems from the experiences gained in working directly with a large industrial firm and incorporting interactions with industry into the curriculum. Moreover, to use his 5-year industry experience to enhance the exposure of students to industrial problems. The development of efficient techniques for simulating large-scale systems holds the potential for significant contributions to the design, control, and analysis of these systems. The principal investigator has previously developed an impressive body of work in the area of importance sampling for queueing and reliability problems. The research builds on this work, and there is strong potential for incremental gains. Moreover, the educational plan promise s to bring many of the results of the research back to the classroom.
