News Archives

PLFS: A Checkpoint Filesystem for Parallel Applications

September 10, 2010

  • Date: Friday, September 10, 2010 
  • Time: 12noon — 12:50 pm 
  • Place: Centennial Engineering Center, Room 1041

John Bent
Researcher Los Alamos National Laboratories

Parallel applications running across thousands of processors must protect themselves from inevitable system failures. Many applications insulate themselves from failures by checkpointing. For many applications, checkpointing into a shared single file is most convenient. With such an approach, the size of writes are often small and not aligned with file system boundaries. Unfortunately for these applications, this preferred data layout results in pathologically poor performance from the underlying file system which is optimized for large, aligned writes to non-shared files. To address this fundamental mismatch, we have developed a virtual parallel log structured file system, PLFS. PLFS remaps an application’s preferred data layout into one which is optimized for the underlying file system. Through testing on PanFS, Lustre, and GPFS, we have seen that this layer of indirection and reorganization can reduce checkpoint time by an order of magnitude for several important benchmarks and real applications without any application modification. The full paper, which was a best paper nominee at SC ’09, can be downloaded at:

Bio: John Bent 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 school. Wisconsin accepted him and seven years later, John finished his dissertation about data-aware batch schedulers. Since then, John has spent the last five years working on parallel storage systems at Los Alamos National Labs. PLFS grew out of that work.