Composite Synchronization in Parallel Discrete-Event Simulation
Speaker : David M. Nicol, Department of Computer Science, Dartmouth College
Time : Friday, April 6th, 3:00 pm - 4 pm.
Location
:
Stuart
Building Room # 111 (SB 111)
Synchronization has long been a topic of study when considering the use of parallel computers to execute discrete-event simulations. Among "conservative" techniques one finds both asynchronous (local) and synchronous (global) approaches. Synchronous approaches are simple and scalable, but are at risk when the worst case synchronization need is much worse than the average case. Asynchronous approaches apply synchronization logic only as often as needed locally, but are at risk when many elements have high connectivity.
Composite synchronization is a new approach that automatically, dynamically, and optimally tailors a composite of synchronous and asynchronous techniques in the same model. In this talk we develop the theory of composite synchronization, prove optimality results, and demonstrate empirically that the technique automatically finds the "sweet spot" composition that optimizes performance.
Short bio of the speaker:
David M. Nicol is Professor and Chair of Computer Science, at Dartmouth College. His research interests include parallel processing, and computer modeling/simulation; he is Editor-in-Chief of the ACM Transactions on Modeling and Computer Simulation, and co-author of the text "Discrete-Event Systems Simulation". He holds a B.A. degree in mathematics from Carleton College, M.S. and Ph.D degrees in computer science from the University of Virginia.