When talking about vCenter Server, VMware View administrators frequently ask questions about the scalability differences between desktop and server workloads. vCenter Server 5.0 has a documented scalability of 10,000 virtual machines. However, VMware View 5.0 has a documented scalability of 2,000 desktops. As of VMware View 5.1 few things changed, as noted below.
VMware Architecture Guide defines the following: “With vCenter 4.1 and 5.0, each vCenter Server can support up to 10,000 virtual machines. This support enables you to have building blocks that contain more than 2,000 View desktops. However, the actual block size is also subject to other View-specific limitations. If you have only one building block in a pod, use two View Connection Server instances for redundancy.”
That is great news, however because of the picture demonstrated below in the VMware Architecture Guide, and because of previous bad experiences with vCenter Server tasks failing to complete because of insufficient resources, administrators and architects in general have start designing VMware View solutions with no more than 2,000 desktops per vCenter.
There is also an old tale that VMware escalation and support services will not support environments beyond 2,000 virtual desktops. I know that because I used to work for VMware PSO and heard many customers mentioning that to me.
Well, the reality is that administrators and architects are able to scale vCenter Server beyond 2,000 desktops – “ as long you know exactly what you are doing “. There are a number of VMware View variables that will affect scalability and performance of vCenter Servers, such as the number of operations that VMware View is executing. Those operations could be related to how often virtual desktops are refreshed or re-composed. As an example, a desktop pool set to refresh on logoff where users only logoff at the end of the day will generate much less operations than a desktop pool where desktops are refreshed every 45 minutes.
Other variables that will impact scalability and performance are related to the underlying infrastructure, such as amount of RAM, storage latency, CPU consumption etc.
All those variables together create a very hard to predict environment, and that is one of the reasons for the 2,000 desktop limit or tale. Another reason is related to what VMware QE team has effectively and successfully tested before the product launch. Most of the times software is tested under certain conditions, not all conditions.
May I increase the number of desktops being served by a single vCenter Server? – Yes
May I change the size of the View Pod to accommodate more than 2,000 desktops? – Yes
May I have a 32 node NFS cluster being supported by a single vCenter Server? – Yes
Will VMware provide support to my environment? – I have no authority to say yes or no , but there are indeed customers running VMware View with vCenter Server supporting more than 2,000 desktops.