ESXi 4 VM Performance

[Dandav]

Installed a new IBM x Series 3650 server for a client. The server has 2 x Xeon Quad Core Processors, 12 GB RAM, 3 x 146 GB SAS 15k drives. Installed ESXi 4 using defaults. Installed 2 Windows Server 2008 Std 32 bit servers as VM's. One is a domain controller, the other is a Terminal Server. Approx. 30 users on the network. The client complained that performance of their Line of Business App is worse than on the old server. Initially had 2 processors per VM and 4 GB Ram each. Increased processors to 4 each with no improvement in performance. The application is a FoxPro based program. I am new to VMWare and don't know what to try to improve performance. We talked to the application provider prior to installing the new server and they recommended the configuration. The old server was 64 bit Windows 2003 but the app provider does not support 64 bit anymore. Any suggestions would be greatly appreciated. Thanks

[a_p_]

Without knowing the details of the configuration, I think the issue could be related to the HDD configuration and the RAID controller.

- Which RAID controller do you use?

- Does it have a BBU (battery backed write cache)?

- Did you configure the logical volumes on the RAID controller to "write-back"?

- How did you configure the disks? RAID1 with a standby disk or RAID5?

André

[Dandav]

Thanks for the reply.

Raid 5, the controller is an IBM ServeRAID M5014 with BBU installed. I didn't make any changes to the controller so I'm not sure about "write back" default settings.

[a_p_]

Raid 5 ...

Even though RAID5 with 3 disks is no optimal configuration for disk performance (for RAID5 you should have at least 4 HDDs), I think this may not be the issue since you have the 15k SAS disks.

...the controller is an IBM ServeRAID M5014 with BBU installed.

That's ok. It's on the VMware HCL and BBU is absolutely neccessary for ESX(i), because the hypervisor does not do software based disk caching..

I didn't make any changes to the controller so I'm not sure about "write back" default settings.

You should check this and make absolutely sure "write back" is enabled for your volumes. The difference between "write through" and "write back" is huge! When a logical volume is created, the default setting (at least for the controllers I configured) is set to "write through" and has to be set to "write back" manually. To make sure you don't loose data, you should also enable the setting to automatically fail back to "write through" if the battery fails.

André

[Dandav]

I will be going onsite Tuesday morning so I will check that. I presume I will need to boot to the Array config utility. If the change is made, is there anything I need to do in VMWare? Thanks again. I'll update Tuesday.

[a_p_]

If the change is made, is there anything I need to do in VMWare?

No, the change is internal to the RAID controller and transparent to ESX(i).

André

[Baggerz1]

How many other vms on the box and how many CPUs do they have?

What used CPU do they have? Ready and wait time?

[Dandav]

Thanks for the help. I checked the RAID controller and it is operating in "Write-Back" mode. After working with the software vendor, we moved the database onto the Terminal Server where the client software is installed.and it seems to run fine. The problem is the application and not VMWare. Thanks again for your suggestions.