We are considering doing a View pilot and I have a few questions. I am already running vSphere 4.1 with 1 vCenter server. The pilot would be about 20 systems and we'd get the Premier licenses.
1) Do I need a separate vCenter for a View implementation?
1a) If so, does it need to be a physical server, or would a VM be okay?
1b) I would probably get a 10-pack bundle and 10 addon licenses, so I'll have licensing for an additional vCenter if needed.
2) I already have a FC SAN for my existing VM infrastructure - is it okay to use this same SAN for View?
2a) The drives are all FC drives so I'm thinking performance should be okay.
3) I know I'll need a dedicated ESX physical server for View clients. For standard Windows 7/Office usage about how much processing power and RAM are needed?
3a) I've looked through some information at vmguru.nl (great site by the way) and I think I have a pretty good idea, but more input from people who are already using View in a production environment would be appreciated.
4) We are looking at the Samsung NC240 PCoIP units for our clients, does anyone have any production experience with these units?
1: For a POC it should be ok to utilize the same vCenter. If you move it into production with a fairly large number of machines I would create a dedicated vCenter instance.
1A: Both virtual and physical configurations are supported.
2: Shoud be ok to utilize the same SAN. As far as the server and desktop workloads sharing the same disk. I guess the bottom line is to make sure you have enough IOPS to handle the VDI workload. I think storage performance is a big part of VDI performance and your pilot will fail if the user experience is bad.
3: We use 1vCPU and 2GB memory for our Windows 7 and Office 2K7. The best thing to do would be to measure the performance of your physical machines to get a good idea.
3A: Another great site is http://myvirtualcloud.net/. There are a lot of sizing tools and in depth articles on how View works available on the site.
4: I don't have any experience with those.
Regarding ESX host CPUs - does it make sense to size a View deployment around CPU cores or GHz? I haven't seen much regarding this so maybe it isn't a consideration, or there hasn't been much testing done...?
I'm thinking number of cores is more important since each client could conceivably have access to its 'own' cpu if you have enough cores...or is this not how things work? So having more cores would be better than having faster CPUs for a View deployment...?
2 x 3GHz quad core CPUs = 8 cores available and 24GHz pool - 20 view clients might contend fairly often for CPU (8 cores)
4 x 2GHz quad core CPUs = 16 cores available and 32GHz pool - 20 view clients would contend much less often for CPU (16 cores)
I'm not sure - I seem to remember something regarding best practices when setting up a View environment to have View separated from the server environment by containing it within another vcenter instance (possibly multiple vcenter instances depending on View environment size).