[Gluster-users] Looking for use cases / opnions

[Thomas Wakefield]

Everyone, thanks in advance.

We are looking to add a large filesystem to our compute facility at GMU.  We are investigating if Gluster can work in a University setting for some HPC work, and general research computing.  Does anyone have use cases where Gluster has been used in a university setting?

The idea we have is to build a 2-3PB GlusterFS system.   We would probably use commodity 1u servers with 60 drive JBODS attached, from a reputable vendor (Dell, HP, etc).  Is 60 the maximum number of drives that’s recommended per server, or could we use 2 jbods per server?  Everything will be connected to a 10GB network.  Clients are mostly linux.

Are there ways to minimize the latency for stat commands like ls and du?  We have some previous experience with Gluster and found the stat performance to be sluggish frequently.  

Thanks,

Thomas