Today VMware unveiled several major improvements in Cloud computing. This was a good webcast, and it appears a link is available here http://www.vmware.com/cloud-infrastructure/business-critical-applications.html to watch the re-run if you are interested.
CEO Paul Maritz and CTO Steve Herrod shared some very cool improvements and facts. Virtualization of Business Critical Applications is on the rise. Between January 2010 and April 2011, virtualized instances of SAP grew from 18% to 28%; MS SharePoint from 53% to 67%. Other critical apps like Exchange, SQL and Oracle all showed growth too. In order to allow for additional growth, the new release (vSphere 5) will support 32 virtual CPUs per VM (up from 8 in vSphere 4) and memory support will increase to 1,000GB per VM (up from 256GB in vSphere 4).
Then Rick Jackson, Chief Marketing Officer (CMO) came on stage. I feel sorry for this guy, because he really stole the show — and nearly spoiled the huge announcement. He announced that the restrictions around cores per processor and physical memory limits per host were going to be removed – simplifying licensing. Part of the justification was improvements in hardware that weren’t conceiviable even a few years back. This is a change for the better, but shortly after things started to go bad. The next slide made reference to a new vRAM pool, which didn’t go into a lot of detail. Twitter and blogs have been a flurry for the last ~12 hours going into this issue. I really like this description/run down of the issue — along with real world implications: http://lonesysadmin.net/2011/07/12/the-five-stages-of-vmware-licensing-grief/
My concern is much more simple. The new vRAM pool model allows for 48GB of RAM to be allocated to virtual machines per socket license (assuming Enterprise Plus Edition). Lets suppose I have some awesome hardware and a business need to reach the new configurable maximum of 1,000GB of RAM on a single virtual machine. This would require that I have around 21 socket licenses to meet my single virtual machines memory demand. While 1,000GB of RAM is now technically possible, it may not be financially possible for many organizations.
I understand the approach — and it is really no different than past licensing models — those who use the higher end features pay more. I also understand the outrage — with the improvements in hardware in the last couple of years, it is now possible to achieve crazy consolidation ratios we couldn’t have conceived in the recent past. In my opinion, this license change may prevent some organizations from moving to the new platform.
I’m looking forward to hearing the VMTN Communities Roundtable Podcast tomorrow…I’m pretty sure a discussion on this topic will break out within the first few minutes.
I didn’t want to write another post on this issue, but VMware announced changes to the vRAM allocations today. This change helps a lot with the monster VMs as well as customer perception of the license changes. Here is a link to the post: Changes to the vRAM licensing model introduced on July 12, 2011