John Siirola, Senior Member of Technical Staff
Exploratory Simulation Technologies
Sandia National Laboratory
“Current Trends in Parallel Computation and the Implications for Decision Support Tools”
Monday, November 2nd, 2009
10:00-11:30 a.m.
FRNY 3059
Abstract: In recent years the historical increase in processor net serial instruction rate has slowed dramatically, with processors first reaching an effective upper limit for clock speed and now approaching apparent limits for microarchitecture efficiency. Current trends in processor development suggest that future performance gains will occur primarily through exploitation of parallelism. While historically many decision support tools have leveraged serial computing, for these tools to continue to leverage the “free” advancements from computing technology in the future, the decision support toolset will need to embrace the use of parallelization.
This talk will survey some of the current and emerging parallel processing environments and present several ongoing efforts for exploiting parallel processing within the context of decision support. While by no means comprehensive, I will focus on three representative efforts: the development of parallel branch and bound algorithms for single-objective optimization, distributed collaborative optimization environments for multi-objective optimization, and massively multithreaded parallel discrete event simulation. Each of these applications leverages distinctly different parallel environments from multi-core to massively parallel to specialty parallel systems in order to provide key capabilities within a general-purpose decision support toolset.
Bio: John Siirola received a BS in chemical engineering from Purdue University in 2000. While there, he participated in the Cooperative Education program with Air Products and Chemicals in Allentown, PA, where he focused primarily on process simulation and computer-aided design, including the development of batch process scheduling, configuration management and on-line data reconciliation applications. He went on to obtain his Ph.D. in chemical engineering from Carnegie Mellon University in 2005. His research focused on developing agent-based methods for non-convex mixed-integer multi-objective optimization using distributed computer clusters.
John joined Sandia National Laboratories in 2005 and is currently a Senior Member of Technical Staff in the Exploratory Simulation Technologies department. His research activities focus on the development of agent-oriented software applications for simulation and optimization of complex network systems with applications in logistics networks, infrastructure systems, and regional markets. Current projects include developing hybrid modeling environments for capturing continuous, discrete event and agent-based paradigms, and developing frameworks for implementing both agent-based and non-agent-based hybrid optimization algorithms.
Posts
