Maximum numbers of Vcpu in single ESX Host
Posted by craig 
- on February 28th, 2009 in Server, Virtualization | 6 Comments »
If we refer to the current version which is ESX 3.5 u3, the maximum number of Vcpu per ESX server is 192 per ESX Servers. Personally, I think the number of Vcpu per ESX servers is too minimal. Imagine if we do run a servers with 4 or 8 physical CPU sockets and we consolidate 40 : 1 Physical server in our virtualization environment, we will hit to the bottleneck on maximum numbers of Vcpu per ESX servers but not due to the CPU consumption.I hope VMWare should revise the configuration to greater number or 256 Vcpu per ESX servers.
The reason of virtualizing today, is really not about sacrifice performance. In our environment, we have most VMs running with 4 Vcpu as the result of customer compare VM to Physical Servers to ensure the performance is sustainable or improve when they switch to VM. When the VM itself is idle or less busy, it will not consume too much resources on the Physical CPU cores. Next release of ESX will go for 8Vcpu per VM. The number of Vcpu been assigned to the VMs which require extra CPU resources will increase too.
Updated: Configuration Maximums for VMware Infrastructure 3 – Number of Virtual CPU = 192
Tags: CPU, ESX, Processors, vcpu, Virtualization, VMware
6 Responses
I guess have to be a little patient hopefully it will be address in the coming release
If you are currently hitting this max on a non-vdi environment than you need to change your provisioning. By the way the current max is 192 and not 128. And my guess is this will go up considering the fact that 8 vcpu’s will be available in the next version.
again keep in mind that using multiple vcpu vm’s as standard will make scheduling much more difficult.
Hi Duncan, I am little confuse about the max number here, according to the document I downloaded from vmware, it show 128 Vcpu per servers. Not sure is the document error. Maybe you can help to clarify on this.
Personally I have to say that different environment have different way to manage. We serve multiple timezone in single ESX servers, result as the busy hours for each VM will be different as it provide flexibility for my case to allocate more resources without impact the rest of the VM on same host. Just my 2 cents to share
Revise = Max Vcpu in single ESX server to 192
I find the limit OK so far.
What surprised me is your statement: ” have most VMs running with 4 Vcpu as the result of customer compare VM to Physical Servers to ensure the performance is sustainable or improve when they switch to VM. ”
Mate you must educate your customers, this is not a 1:1 scenario, 1 physical server wit x CPU=1 virtual machine with the same number of vCPU.
Read more about how VMware co-schedules physical processors …
actually, if you refer to my comment on this, it does describe that the VMs we run are serving multiple regiosn which contain, ASIA, EURO and NA in same ESX servers. In the case, we could actually allocate more resources to the VM as the production hours for each region is different. Of course is not 1:1 comparison on the Cpu, we only provide to the system which really need it.