Like v4 the licenses will be by per physical processor. The difference is where v4 had a physical limitation on memory and processor cores. v5 does not. However, v5 has adopted what may very well cause people from upgrading from v4 to v5 and that's the vRAM entitlement setting on each processor that is licensed. This entitlement (24GB vRAM/proc for Standard, 32GB vRAM/proc for Enterprise, and 48GB/proc vRAM for Enterprise Plus) causes concerns for large businesses like my customer that I support at my job; and here's why...
Say I have a robustly built hardware environment that consists of rack and blade servers (10 for example) that each has 512GB of physical RAM and 4 processors each. Say all 10 of those servers are centrally managed via vCenter and my licensing model is Enterprise Plus (because I like vDS and Host Profile use)... Under v4 I had few limitations with a physical memory maximum of 1TB per host and a VM vRAM limitation of 255GB. With v5 that goes away but I’m now restricted even more because of the vRAM entitlement.
10 hosts x 4 procs = 40 procs that need to be licensed at Enterprise Plus
40 licenses x 48GB vRAM entitlement = 1920GB vRAM entitlement.
However, I have purchased 5120GB of physical RAM for my environment. Therefore, I’m only able to allocate approximately 27% of my memory (5120/1920=.2667) before I’ve reached this entitlement capacity. Once reached, I have to purchase additional licenses (or upgrade my licenses if I have that option). This drives up front costs significantly in order to allow us to utilize RAM that we purchased. In my example my costs for licensing alone (not factoring in support or vCenter) would be approximately 140K to license all 40 procs (MSRP). If I were running vSphere 4, I would be able to utilize all 5120GB of RAM on my hosts. With v5 I have to purchase approximately 30 lienses
5120 * .75 / 48GB (vRAM entitlement for ENT +) = 30 licenses
That drives my costs up by over 100K.
Thus my concern… While VMware has partnerships with literally everyone (Dell, HP, Cisco, EMC, etc.) one would think that if I’m a hardware vendor I would express my dislike at this limitation that my customers now face and thus reducing revenues on my organization (I know an ugly marketing angle). We’ve come so far from servers back in the early 2000s that had 1-4 GB capacity to servers that can house 1TB of RAM per server. VMware has proclaimed for years about consolidation onto robust hardware and though the vRAM capacity is supposed to be agnostic from the hardware and be presented as a shared pool of vRAM resources (think cloud computing here but on the local infrastructure). The limitations should have been increased or eliminated for those with Enterprise Plus licensing. It’s because of this limitation that concepts like over commitment of memory is no longer needed but with robustly built hardware (like my example) goes to waste (thus organizations that have already purchased this equipment now have wasted money on something they will doubtedly ever use.
Though I’m not the only one with this concern or expressed dislike in this new design feature, I know it’s too late to pull it back as vSphere 5 is being announced at VMworld in Las Vegas in about six weeks. However, I hope the feedback VMware is getting forces this idea to virtually disappear as revision patches are released for the product. This can’t be a good step in the right direction. I like the concept of a collection of resources in a pool but there has to be a better way of tying this in without having to involve licensing models. Here’s the White Paper from VMware on the product as well as a comparable maximum White Paper on vSphere 4.x. Enjoy and let me know what you think...
VMware vSphere 5 Licensing
VMware vSphere 4.x Configuration Maximums
