Any error(s) reported in the vmware.log file regarding access issues with the virtual disk?
What host CPU utilization do you get when you perform a large file copy in your VM?
OK, more information...
The CPU utilization is about 4% to 6% on the host and 2% or 3% on the VM.
In the vmware.log, there are a lot of the following messages:
vmx| SCSI0:0: Command READ(10) took 1.040 seconds (ok)
vmx| SCSI0:0: Command READ(10) took 1.110 seconds (ok)
There are also WRITE messages.
On the other Win2k VM on the same host, the I/O is pretty good. And I don't see the SCSI READ and WRITE messages in the log. I think these messages might indicate the problem.
OK, something is weird. I gave the VM 1 more CPU and hoped the problem will go away. But it didn't. So that I changed it back to single process VM. And the slowness problem went away. I don't see the SCSI read/write messages in the log any more. I'll run more tests tomorrow and post back the result.
This slowness problem has bothered us for several weeks. This VM reboots every night and the problem always exists.
BTW, now the VM shows "ACPI Multiprocess PC" in the device manager. How can I change it back to "ACPI Uniprocess PC"?
1 person found this helpful
BTW, now the VM shows "ACPI Multiprocess PC" in the
device manager. How can I change it back to "ACPI
Check this thread:
url=http://www.vmware.com/community/thread.jspa?threadID=29415&tstart=0How to repeatedly upgrade Windows 2003 to SMP and revert to uniprocessor[/url]
Bad news! The slowness problem doesn't go away. I tested it last night. It was OK. But It became slow again this morning. The VM gets reboot every night. It's possible the OS reboot messed things up. I'll try a power off/on later when the users are away.
Besides the SCSI messages I didn't find any error in the vmware.log. Are these SCSI messages normal?
BTW, The other win2k VM using Buslogic driver doesn't have the slowness I/O problem.
Lars, thanks for the link. I'll check it up.
Thanks for any help.
OK, the other win2K VM also has the I/O slowness problem. After rebooted the guest OS, the I/O was OK. But after let it run for a while, the I/O slowness will occur until reboot it again. Whenever this slowness happens, the vmware.log will show those SCSI Command READ/WRITE messages. Does any one know if these messages are normal or not?
I'll reboot the host tonight and see what happens.
Thanks for any help.
how much memory does your VM have ?
you may open your .vmx file and set mainmem.usenamedfile="FALSE" - power down and re-start your VM and see if this makes a difference! please report!
The host has 4 GB. One VM has 2 GB and the other VM has 1 GB.
OK, I disable memory page trimming, usenamedfile and page sharing. Let's see what happens tomorrow. I'll report back.
Thanks for your help.
OK, the slowness problem is resolved. The Vm's are really flying now. I think set usenamedfile to false help a lot. Thanks Roland.
However, since I also disabled memory page trimmimg and TPS. The host memory usage is really high now. I'm thinking about enabling memory page trimming and/or TPS to release some memory to the host or other VM's. Does any on have comments on these 2 parameters? I did some researches. It looks like disable them will help performance. But how much do them help (a liitle or a lot) is not very clear?
Thanks for any comment.
By the way I run a domain controller inside a VMware Server 1.0.3 and one of the good thing that fixed disk performance was not only to disable the memory trimming option but to disable the write cache over each virtual hard drives.