2

I have 5 servers, 16 threads running on each of them, producing a total of 80 files at the same time, which should finally be available for every other server in the same folder (it is a MPI parallelized Finite Element Software).

What I did so far: Mounting a NFS folder of Server 1 and writing via Ethernet on Server 1. -> 80 Threads want access on 1 hard disk through 1Gbit LAN at the same time. = "Bottleneck"

From our point of view it would be the best if each server writes on its own disk, but the user sees one combined file system, containing 80 files.

We are running Ubuntu 16.04, and we need something like a cluster file system / distributed file system as far as I understand this.

Which would you recommend: (Ceph / Lustre / Gluster ... or others?) - it should be easy to install, configure and maintain - stable - automatically started / initialized at reboot. - the cluster users don't know about this feature and can't repair it if something goes wrong...

0

You must log in to answer this question.

Browse other questions tagged .