Disk IO bottleneck problem on VMWare Windows 2012 VM on VMWare with NAS storage

Discussion forum for Enterprise Edition.
Post Reply
beachard
Posts: 7
Joined: Thu Dec 18, 2008 3:42 pm
Location: Laval, Canada

Disk IO bottleneck problem on VMWare Windows 2012 VM on VMWare with NAS storage

Post by beachard »

Hi everyone,

We are running MailEnable enterprise server on a virtualized Windows 2012 on VMWare ESXi 5.5 with a NAS datastore. The VM is configured with 2 vCores and 6 Gigs of RAM. We are running 1822 email addresses in 220 postoffices. We are also running 10 mailbox with ActiveSync that seams to be very consuming resources on the server.

Twice during peak time, the server became almost unresponsive and we ran into some sort of bottleneck issue where the disk access was hitting arround 100 IOps on VMWare and we were getting latency up to 6000 ms !! The disk was getting so slow that the spools were rising up and the server was unable to process the traffic. The only way to resolve it was to shut down the services and start them one by one slowly.

My provider hosting the VMWare environment tells me that the technology we are using for the Datastore (NAS) was not capable of handling so much IO so i needed to change to fiberchannel $$$$$$$. Lots of money and moving the VM... etc

MY QUESTION :
Do anybody have any suggestion to improve the server performance for MailEnable ???

1. Would it be a good idear to create another VDisk only for the SPOOLS and maybe another one for the LOGS to seperate the IO on 2 or 3 disks ??? Will it help ?

2. Could i create a RAM DISK and map the SPOOLs on that disk ??? (If so, what size is recommended ?? 4, 8, 16 Gig ?

3. Should i increase the memory on the VM ?? Will it help since it appears to be a disk IO problem ??

4. How do i move the SPOOL directory only ??

Thaks for your help guys !

Post Reply