johnny_the_knife
asked on
VMWare Terminal Services Performance
I have the following situation. We have a customer with an IBM x3500 server equipped with 16GB of RAM, 2 Intel Quad Core Xeon E5410 processors and three 7200RPM SATA drives in a RAID five array. The server has CentOS 5 (64 bit) installed with VMWare Server 2.
Two virtual machines are configured:
(a) SBS 2008 with 4GB of memory allocated & 2 Processors
(b) Windows 2008 Terminal Server with 8GB of RAM & 2 Processors
The Terminal server is supporting 11 users, acessing it via Wyse thin clients on a LAN, all running Office 2007. Throughout the day, users notice their terminal services sessions become unresponsive. This happens up to ten times per day. I don't have an exact figure on the amount of time they are unresponsive for, but I suspect it is from 5 seconds to perhaps a minute. Sometimes the sessions continue to respond, but, when the users are typing, there is a delay between pressing keys and the characters appearing on their screen.
I have witnessed the Processor utilisation sit at 100% for short amounts of time, but it seems to average about 60%. Memory usage never exceeds 60%.
I have configured Windows System Resource Manager in Per User mode (this seemed to improve performance over per session mode) on the terminal server which made some improvement, but didn't resolve the problem all together.
I have also configured VMWare Server to only use available physical memory, so it shouldn't page.
Checking the event logs on the terminal server shows only a couple of Microsoft Outlook applications hangs.
There has been some suggestion that reducing the allocated processors for the virtual machine to one might improve performance, but I am hesitant to do this as VMWare advise that this can cause "instability" if done after the virtual machines have been created.
I need to determine why the CPU usage is so high at certain times, and what I can do to give the users a better experiance.
Two virtual machines are configured:
(a) SBS 2008 with 4GB of memory allocated & 2 Processors
(b) Windows 2008 Terminal Server with 8GB of RAM & 2 Processors
The Terminal server is supporting 11 users, acessing it via Wyse thin clients on a LAN, all running Office 2007. Throughout the day, users notice their terminal services sessions become unresponsive. This happens up to ten times per day. I don't have an exact figure on the amount of time they are unresponsive for, but I suspect it is from 5 seconds to perhaps a minute. Sometimes the sessions continue to respond, but, when the users are typing, there is a delay between pressing keys and the characters appearing on their screen.
I have witnessed the Processor utilisation sit at 100% for short amounts of time, but it seems to average about 60%. Memory usage never exceeds 60%.
I have configured Windows System Resource Manager in Per User mode (this seemed to improve performance over per session mode) on the terminal server which made some improvement, but didn't resolve the problem all together.
I have also configured VMWare Server to only use available physical memory, so it shouldn't page.
Checking the event logs on the terminal server shows only a couple of Microsoft Outlook applications hangs.
There has been some suggestion that reducing the allocated processors for the virtual machine to one might improve performance, but I am hesitant to do this as VMWare advise that this can cause "instability" if done after the virtual machines have been created.
I need to determine why the CPU usage is so high at certain times, and what I can do to give the users a better experiance.
Just as MrN1c3 says i would seriously consider going to esxi (free) it would handle the performance much much better. the only issue is you would need to convert the machines to esxi with vmware converter first.
ASKER
Can you provide some more info on the esx conversion. The server we have isn't on the hcl, but it is pretty standard so I can't imagine that will be a big issue.
ASKER
Just a note, that processor utilisation number is within the terminal services virtual machine, not the host OS.
ASKER
Just doing some more checking and monitoring on the problem server. The CPU sits at about 70% most of the time.
What I've notcied is the wmiprvse process consumes between 8 and 55% of CPU resources all the time. Research found this article, I've ordered the hotfix from MS and will apply it over the weekend. Heres hoping this resolves the problem.
If not, then I'll convert to ESX, but I want to avoid that if I can as the VMs are large (one if 340GB), so it will take a LONG time to do.
What I've notcied is the wmiprvse process consumes between 8 and 55% of CPU resources all the time. Research found this article, I've ordered the hotfix from MS and will apply it over the weekend. Heres hoping this resolves the problem.
If not, then I'll convert to ESX, but I want to avoid that if I can as the VMs are large (one if 340GB), so it will take a LONG time to do.
ASKER
MS artical from above post: http://support.microsoft.com/kb/970067/en-us
What kind of Security Software (Antivirus) are you running within the Terminal Services Server ?
ASKER
Sophos with on access scanning disabled.
1.How is your Network Setup (Gigabit Ethernet or Fast Ethernet) ?
2.What kind of NICs is on the Host ? (Gigabit or Fast Ethernet)
3.I agree with other experts in installing ESXi as the baremetal hypervisor which would enable you to isolate any issues related to Cent OS.
4.If your Hardware is not on VMware HCl, then its officially not supported by VMware.
You can use the below mentioned URL which provides you with a list of supported Servers and PCs which can run ESX/ESXi 4.0 even though the same list is not tested and verified by VMware.
URL - http://www.vm-help.com//esx40i/esx40_whitebox_HCL.php#Servers
2.What kind of NICs is on the Host ? (Gigabit or Fast Ethernet)
3.I agree with other experts in installing ESXi as the baremetal hypervisor which would enable you to isolate any issues related to Cent OS.
4.If your Hardware is not on VMware HCl, then its officially not supported by VMware.
You can use the below mentioned URL which provides you with a list of supported Servers and PCs which can run ESX/ESXi 4.0 even though the same list is not tested and verified by VMware.
URL - http://www.vm-help.com//esx40i/esx40_whitebox_HCL.php#Servers
ASKER
It a gigabit nic but running at 100 meg. I'm not sure this is the issue though, as when the ts vm is slow, I can access the sbs vm via rdp at normal speeds.
The VM's Virtual NIC speed is as fast as the physical NIC.
I suggest that you configure speed (100 Mbps) and duplex (Full) manually instead of auto-config at the Cent OS Host level.
Have you thought about migrating to ESXi 4.0 Free Edition ?
I suggest that you configure speed (100 Mbps) and duplex (Full) manually instead of auto-config at the Cent OS Host level.
Have you thought about migrating to ESXi 4.0 Free Edition ?
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
I would not be comfortable with running vmware server in a production environment, as its a hosted virtual environment, and not a true hypervisor; i.e. it is installed on top of windows or linux.
I would seriously consider to switching to ESXI (which is also free). ESXI is enterprise class and will offer you much better performance.
https://www.vmware.com/products/server/faqs.html