Speaker: John Bent
Scientist, High Performance Computing Division
Los Alamos National Laboratory
Abstract: Current petascale supercomputers have been carefully provisioned to provide a computing platform well-balanced for current workloads. Unfortunately, the IO patterns of many of these workloads are poorly matched to the underlying storage systems; for these workloads, achieving even a fraction of aggregate spindle speed can be very challenging. This talk discusses our recent efforts with PLFS: a virtual file system that rearranges workload IO into patterns better matched to the underlying storage system. Current status of the PLFS project will be presented as well as several other near-term related research projects.
Biography: John's first real job was moving chicken on an assembly line to make extra money for college. After graduating with a bachelor's in anthropology from Amherst College, John spent two years as a public school librarian in Palau as a Peace Corps Volunteer. He then worked for a personal injury lawyer for a year while applying to CS graduate schools. Wisconsin accepted him and seven years later in 2005, John completed his dissertation about data-aware batch schedulers. Since then, he has been working on parallel storage systems at Los Alamos National Labs. His research interests include parallel computing, storage systems, data management, checkpointing, and data intensive computing.