Thesis
Computer and Information Science
M.A.S., The Ohio State University, 1978
B.A., Augustana College, 1977
Doctor of Philosophy
Dissertation
Abstract of Dissertation
October, 1995
Abstract
Acknowledgments
I am deeply indebted to the late Professor Yehuda Wallach, my first advisor, for both inspiring me to pursue my degree and providing me with my initial insights into the use of diakoptics to provide parallelism in power systems network applications. To say that Professor Wallach was passionate about diakoptics would have been an understatement!
I am grateful that Professor Wallach accommodated my part-time student status with weekend meetings at his home. These meetings invariably included the gracious hospitality of tea and snacks provided by his wife, Rachel. I worked with Professor Wallach during the Gulf War (Spring semester 1991), and several meetings were interrupted as Yehuda and Rachel anxiously awaited telephone calls from their daughter in Israel to indicate that she was safe after Iraqi Scud missile attacks. After a telephone call confirmed his daughter's safety, Professor Wallach would resume our meeting.
My research confirmed that Professor Wallach's intuition was correct concerning the applicability of diakoptic techniques as the basis for efficient parallel linear solver algorithms for power systems network applications. May this research stand as a testament to his memory.
* * *
There are many individuals in the academic community to whom I am deeply indebted. First, I would like to thank my co-advisors, Professors Geoffrey C. Fox and Sanjay Ranka, for their assistance and guidance during my years of work on my research. As co-advisors, they provided a good balance of interest in my research. Professor Nancy McCracken, the third member of my research committee, provided guidance and much needed support as I attempted to balance a (nearly) full-time job, my academic research, and my family. I would like to thank Professor Tony Skjellum of Mississippi State University and Professor Chis Pottle of Cornell University for their contributions to this research: their comments helped me make notable improvements to this work. Special thanks go to Alvin Leung, a Graduate Research Assistant at the Northeast Parallel Architectures Center (NPAC), for porting software developed for the Thinking Machines CM-5 to the IBM SP1 and SP2 and for running benchmarks of the software on that architecture. Lastly, I would like to express my gratitude to Dr. Paul Coddington, Senior Researcher Scientist at NPAC, for his contributions to this research and for his friendship.
I owe additional thanks to Professor Fox, who as director of NPAC provided me with the infrastructure, an office, a workstation, travel funds, and access to parallel computer resources, to pursue my research. This support was funded in part by Niagara Mohawk Power Corporation, the New York State Science and Technology Foundation, the NSF under co-operative agreement No. CCR-9120008, and ARPA under contract #DABT63-91-K-0005. I am also grateful to have had the opportunity to professionally associate with the many fine intellects that Geoffrey has assembled at Syracuse University.
* * *
I gratefully acknowledge the financial assistance and the time provided me to work on my research by my employer, The MITRE Corporation, through their Advanced Graduate Degree program.
* * *
Finally, I wish to express my special gratitude for the many years of support from my loving wife, Ann, and my super children: Yuri, Zachary, Naomi, and Abram. I hope that my children are as proud of me as I, at the age of thirteen, was of my father when he received his high school equivalency diploma.
Ann, I love you so very much.