simple tests

[lightp2]

I have some free time but no budget
I did 2 small tests, hopefully means something to other admins.

Host : VMware ESXi 4.0 Update 1
CPU : Xeon 5405
RAM : 8GB

VM : Windows 7 RC1 patched
vCPU : 2
RAM : 3GB
SW : OpenOffice 3.1.1
SW : Microsoft Security Essential
SW : Adobe Acrobat Read 9.3
SW : Microsoft SQL Server 2008 Trial
VM Disk : One virtual disk 128GB (Thin I think) on remote NFS Server

Test 1 inside VM : 10 minutes after boot stablized
Microsoft Security Essential Quick Scan : 38 sec

Test 2 Booting VM
I use console to start vm
1. 1st 10sec : loading of about 30MB over network
2. Next 10sec : low to no traffic.
3. Next 16-18sec approximately 378MB to 380MB load over network
4. next 1 to 2 minutes various windows services started and stopped series of continuous but low network traffics.

VMware ESXi is a good software. I think the 100 dollar online training is pretty reasonable for getting access to such good freely available software. I wish VMware would sell T-Shirts though

 

[SpaceHonkey]

Change vCPU count to 1. You should never use more than 1 vCPU unless you have a multithreaded application that absolutely requires it. 1 vCPU will be faster.

 

[lightp2]

Hi SpaceHonkey, Thank you for your pointer.

1. I set numvcpus = "1" inside the VMX file.
2. I shutdown and restart ESXi Host
3. I shutdown and restart NFS Server.
Step 2 and 3 to prevent filesystem buffer cache from affecting the results.
4. I use console to start VM

If above steps are wrong kindly advice.

Host : VMware ESXi 4.0 Update 1
CPU : Xeon 5405
RAM : 8GB

VM : Windows 7 RC1 patched
vCPU : 1
RAM : 3GB
SW : OpenOffice 3.1.1
SW : Microsoft Security Essential
SW : Adobe Acrobat Read 9.3
SW : Microsoft SQL Server 2008 Trial
VM Disk : One virtual disk 128GB (Thin I think) on remote NFS Server

Test 1 inside VM : 10 minutes after boot stablized
Microsoft Security Essential Quick Scan : 35-37 sec

Test 2 Booting VM
I use console to start vm
1. 1st 11sec : loading of about 31MB over network
2. Next 8sec : low to no traffic.
3. Next 27sec approximately 382MB load over network
4. next several minutes various windows services started and stopped series of continuous but low network traffics.

So I think between 2 vcpu and 1 vcpu
1. Boot VM test the major loading is 16-18s against 27sec
2. Microsoft Security Essential Test 38sec against 35-37 sec. Since this is un-controlled visual time count, can say they are about the same, implying in this case the I/O is more of the determining factor.
3. I have some simple SQL test jobs. I did not test this. I like 2 vCPU for SQL Server but not sure whether my simple software load warranted it. Will report if anything interesting.

 

[lightp2]

The result is indeed as suggested by vCPU=1 recommendation.

Performed a small test run.

ESXi 4.0U1

VM-1 WinXP-SP3
SQL Job. (read file data,convert and push to db server)

VM-2 Win7 RC1 Patched
SQL Server 2008 Trial

Case 1 : VM-1 with vCPU=2
Test run approximately 3:30 minutes

Case 2 : VM-1 with vCPU=1
Test run approximately 3:03 minutes

The test run was performed with only these 2 VM on single host.

The vCPU count was also changed on the VM-2 between 1 or 2. It has minimal difference for my small test run.