Hey All,
I'm running ESXi 3.5U4 on a HP DL360 G3. The Specs are as follows:
CPU - 2x 2.8Ghz Xeon
RAM - 4GB
Disk - 2x 147GB 10K RAID1
All i have is one windows server, and i figured it would be at least as fast as if it were running it on the bare metal, but it is extremely slow, and using 100% of the CPU at times... Even going to the control panel peaks it to like 5000Mhz.
Any ideas how to better tune this?
Crank it down to 1 vCPU and see if performance improves.
How many virtual CPU's do you have assigned to the VM? What version of windows server are you running? Is this a new build of the VM or a P2V conversion? Can you provide any additional information?
The VM is 4vCPU's, with 4GB of RAM. Its Windows Server 2003 Enterprise, and it is a new build.
Crank it down to 1 vCPU and see if performance improves.
I'm running ESXi 3.5U4 on a HP DL360 G3. The Specs are as follows:
CPU - 2x 2.8Ghz Xeon
RAM - 4GB
Disk - 2x 147GB 10K RAID1
The VM is 4vCPU's, with 4GB of RAM. Its Windows Server 2003 Enterprise, and it is a new build
You appear to have allocated all of your installed RAM to your VM. You need to keep a bit of Host RAM in reserve for ESX to run in or else you will run into swapping. Reduce the amount of RAM allocated to your VM.
Also, as a previous poster stated, you should reduce the number of virtual CPUs as well. I suggest starting with 1 vCPU and 1024 MB of RAM as a start and go from there.
Also,
Have a look at what processes are using all the processor resource.
I cranked it down to 1 vCPU and it runs much faster now.
So being this is a dual processor single core server, I'm pretty much stuck with only 2 vm's on each host before I see a performance hit?
There are lots of reasons why you would experience unfavorable performance on that server.
a) thats 6+ year old technology
b) your ram allocation is way out of reach
c) always start with a single vCPU, upgrade as performance needs demand it
Creating virtual machines is not a "slicing" function by nature. You can create resource allocations that always leave a prescribed portion of CPU/MEM available for a particular VM or Group of VMs, but thats as far as any slicing notion goes (this is not the same as Mainframe partitioning).
Your VM would perform much better with a single vCPU, 2GB ram. However, in that capacity, the 4GB RAM amount in your system leaves very little room for expansion. Can your system support more ram? Even if it does, can those P4-spec Xeons handle any additional load? Are you just doing this work for testing?
So being this is a dual processor single core server, I'm pretty much stuck with only 2 vm's on each host before I see a performance hit?
Not necessarily. But configuring a VM with more than 1 vCPU makes it harder for ESX to schedule time for the VM because it has to wait until more than 1 core is free before it can continue. If you stick with 1 vCPU VM's you should be able to run several VMs on there. You will most likely run out of RAM before you run out of CPU.