[HPDD-discuss] Disk provisioning for Lustre

I am setting up a small Lustre file system to use with our clusters. It is based on Dell R415 (MDS) and Dell R515's for the OST's. Each R515 has 14 disks. Two are configured as RAID 1 for the OS. The other 12 are 2TB Near Line SAS, currently configured as a single RAID 6 array. There are four R515's as OST's. They are connected to a DDR2 Infiniband network, along with all the cluster nodes. All servers are currently set up with Lustre 2.3. I am looking for advice on how the RAID setup should be configured in terms of stripe size, read caching, read/write policy, etc. Right now I have a test file system configured with each OSS running with a single OST configured from the 12 disk as RAID 6. Is there an advantage to partitioning these in to smaller virtual disks and having multiple OST's on each OSS? -- Ray Muno Information Technology Supervisor University of Minnesota Aerospace Engineering and Mechanics Mechanical Engineering 110 Union St. S.E. 111 Church Street SE Minneapolis, MN 55455 Minneapolis, MN 55455