Preparing for Exascale: Performance Analysis Research at LLNL

Todd Gamblin
Center for Applied Scientific Computing
Lawrence Livermore National Laboratory

June 11th, 13:30--14:30
Seminar room W8-809, Tokyo Institute of Technology, Oookyama Campus.

Abstract:
Exascale machines that are expected to arrive by 2018 will have between 500 million and 4 billion concurrent threads. Lawrence Livermore National Laboratory (LLNL) is currently preparing for Sequoia, which will arrive in early 2012 and have 1.5 million cores and 6 million threads. As current methods simply will not scale to those extreme levels of concurrency, we must develop a new generation of scalable tools to manage, to gather, and to analyze the massive performance and debugging data that we will encounter as we attempt to use these systems effectively. In this talk, I will discuss LLNL research projects on scalable tools and our strategy for enabling exascale application development.

Speaker bio:
Todd Gamblin is a Postdoctoral Scholar in the Center for Applied Scientific Computing (CASC) at Lawrence Livermore National Laboratory (LLNL). As a member of the Application Development Environment and Performance Team (ADEPT), he works on LLNL's Exascale Computing Technologies (ExaCT) project and the SciDAC-sponsored Performance Engineering Research Institute (PERI) project.  His research interests include in-situ performance monitoring and analysis, large-scale simulations using MPI, load balance, scalability, and run-time systems.

Todd received his Ph.D. degree in Computer Science from the University of North Carolina at Chapel Hill in 2009 and his M.S. in 2005.  He earned his B.A. in Computer Science and Japanese from Williams College in 2002.  He is a member of both IEEE and ACM.

Contact: Naoya Maruyama (naoya at matsulab.is.titech.ac.jp)
GSIC Seminar 2010. Co-sponsored by Global COE CompView (CompView Seminar10-02) and New IT Infrastructure for the Information-explosion Era, MEXT Grant-in-Aid for Scientific Research on Priority Area