|
|
Thomas Sterling
Faculty Associate
in the Center for Advanced Computing Research
Professor of Comptuer Science in the Center
for Computation & Technology at Louisiana State University
Contact Information:
Caltech |
Louisiana Sate University |
|
|
|
|
Research Objectives and Directions
The achievement of high performance computing depends on high speed and capacity technologies, and the exploitation of parallel computing structures. Ultimately, practical computing capability is bounded by cost, whether measured in dollars, power consumption, or size. Delivered performance is a function of peak capacity and efficiency which for many parallel systems is itself determined by latency, overhead, contention, and starvation. My research objectives have been to devise 1) execution models that expose myriad forms of parallelism, 2) architecture structures that minimize these sources of performance degradation, 3) dynamic adaptive resource and task management mechanisms that further mitigate or hide the effects of such factors, and 4) software strategies that supervise application to system interfaces. To this end, I have engaged in research of a diversity of physical and abstract structures, often of my own devising. In addition, I am interested in the exploitation of highly replicated structures to provide dramatic reliability advances through dynamic graceful degradation.
Major Research Activities
Hybrid Technology Multithreaded Architecture (HTMT)
JPC4-4 (4th Joint PC Clustered Computing Conference