slow nfs performance

Hey volks,

This is my first post, so, please do not judge me when i forget some basic stuff like logfiles, the reason is i have no idea which one are needed by the Community to support me.

I have the following setup:

2x 1TB (S-ATA) HDD in a 4GB DualCore Workstation
2x 1Gbit Interface (Bonded with balance-tlb)
3x ESXi 5 Hosts with 2x2GBit Interfaces (attached to the virtual machines, load-balanced)

I have created two volumegroups and assigned each the max on available Space (4GB Cache excluded)

I created NFS shares and mounted them to a ESXi 5 Machine without any problems.

The performance decrease drastically then more machines where added on the Share. Means: When i install one machine speed is (based on my feelings) ok, when i install a second one, speed is a bit slower, but ok, when i add the third machine, and a fifth one as example, the speed goes down and working with RDP is not possible with having no long wait times. When looking into the taskmanager the CPU/MEM is idle - so, not a performance problem on that side i think.

I'm new to NFS mounts, maybe i did a initial config error? Any suggestions how many virtual machines should run without problems with my configuration?

As i have tested a lot of configurations now i need your assistance on that...

Greetz,
Matthias

P.S. What are good ways to test the performance and find bottlenecks?

Have you tried using iSCSI instead of NFS ? You can use MPIO, and get more through put this way.
You don't mention drive speeds either. What kind of drives are they, and are you using software RAID, or not?

Hey!

Thanks for your response. Unfortunately i don't know exactly what drives in it, the only thing i know about is:

SAMSUNG | 1TB | 7.200 RPM | ?? CACHE
SEAGATE | 1TB | 7.200 RPM | ?? CACHE

I don't have ISCSI setup done yet, i will give them a try if this would increase the performance.

Normally my goal was to have around 10-15 machines to run on this kind of Setup (NFS/2x1Gbit) - is this possible per design from a storage perspective??

I created the both NFS pools with each 4GB of cache-space, is this ok? (hole partition size is 900GB each).

Greetz,
Matthias