Where is my NFS data physically located?

Introduction

CSAIL has NFS servers in two different data centers. It is intended that this be mostly transparent to you, but if you are doing intensive computational work on your data, such as with a GPU cluster where you need to read data as fast as possible to keep the GPU from stalling, you may need to consider where the data is in order to request the right compute resources. If you are using our Slurm batch queueing system, you can explicitly request resources in either location.

How to check

The easiest way to check which server is located where is to use the df command to show how much space is free. In the first column of the command’s output, for each NFS mount, it will display the name of the NFS server and the name of the filesystem as that server knows it (which is different from the public name used throughout the CSAIL Linux environment).

For example:

Filesystem                                  1K-blocks        Used   Available Use% Mounted on
nfs-prod-12.csail.mit.edu:/export/scratch 53687091200 29736603008 23950488192  56% /data/scratch

This says that /data/scratch comes from the server nfs-prod-12 and on that server it has the name /export/scratch. (Server and remote path names subject to change.)

NFS servers located in Stata

All servers are located in 32-341 except as indicated.

NFS servers located in Holyoke

All servers are located in OC40-250.