It all starts with a quick overview of your local journaling file system. Next on to distributed file systems NFS and CIFS ( aka SAMBA ). With that background, the leture will then turn to High Performance Distributed File Systems GPFS, Luster, HPFS and Ceph. The talk will focus on the area of real production systems. What aspects to keep in mind when designing and deploying very large file systems. Since the talk will start small and build up, everyone should be able to learn something and thus appeal to a majority of our community.

Joseph Apuzzo has been with IBM’s HPC effort for 17 years. During that time he spent 15 years supporting GPFS. General Parallel File System ( aka GPFS ) is IBM’s high performance file system which is used by large and small companies around the world to store billions of files in peta-bytes of storage. Joe has taught GPFS skills in Italy, Korea and the United Kingdom as well as here at IBM Poughkeepsie.

Talk Slide Deck:
FS from Simple to Distributed-HP.pdf
FS from Simple to Distributed-HP.odp

Lightning Talks

  • Google Project Fi - Sean
  • Maker Faire Poughkeepsie Recap - TBD
  • Currently looking for volunteers. Sign up if you’d like to give one.