A bit of background to this issue. We have a SAN datastore for a number of servers. Originally we had a single HP Proliant ML 350 G6, connected to the SAN via NFS. The SAN served a number of Linux guest servers which themselves were running NFS to provision shared folders between the Linux OS's.
The problem we found was an inherent slowness in the NFS serving file requests, which caused time stamp issues between shared files. We initially moved the guests from the SAN to the local storage of the HP server. But again, there was still a slowness to NFS running between the the Linux guest OS'S. There isn't standardisation between the Linux guests, so they are a variety of Red Hat, Ubuntu and Fedora. The server is well spec'd with 48GB RAM, Intel Xeon E5530 with two quad core processors, quad 1GB NIC & 15k 300GB SAS drives. The utilisation of all of these resources is well within the thresholds.
The question I have though, is there an optimal configuration for running Linux hosts with share NFS storage on VMware. We are about to invest in some new hardware for these guest OS's. If there is a particular server that handles this better or whether we should buy a server with two disk arrays and controllers and have the NFS run off one & the other configured for the ESXi hosts. Or should we upgrade to ESXi 5.1 if this manages NFS any better. I have seen numerous posts about similar issues, but I would love to know if anyone has nailed a solution or at least got reasonable performance. This seems to be an issues with VMware and NFS run from within the guest OS, but I can't see why.
Thanks,
R.