Hi, Jim Blakley, first post: I'm Director of Data Center Virtualization in Intel's Digital Enterprise Group. My role is to figure out what Intel can do to make virtualization real in end user environments. The only way to do that is if you help. At Intel, we're mostly technology wonks so we try to find a technology hammer for every problem. I love hammers too but also want to find what we need to be doing to get the industry to solve the big problems and the little problems.
I've talked to lots end users, ISVs and systems providers. Our virtualization labs and our own IT groups here in Intel have a reasonable view of some of the big issues based on hands on work (at least we like to think so).
Server virtualization is very real today. Test/Dev usages have proven the technology as viable and consolidation usages are in production for simple, low utilization applications. People are loading up 10-20 VMs per server in real deployments. You can also see some of the more interesting usage models that people are doing here. But, ...
Today, not all applications are candidates for virtualization. IO and Memory limitations are the first barrier. Even medium sized apps like Exchange and Citrix Servers seem to be too IO and/or memory constrained to be virtualized. (Let alone big ones like ERP and large databases). Some applications need greater security or reliability than current virtualized solutions allow. Some apps are likely never to be virtualized because doing so is simply too costly and complex and the return is not there.
Question for the community: What apps can't or won't you virtualize today? Why?
Has anyone virtualized (in production) any of the IO/Memory intensive apps? Which ones?
The ROI for doing that first phase of server consolidation is obvious: fewer servers in the data center, less power, less space, less hardware to buy, fewer network ports. The ROI for moving to virtualize more complex and more loaded applications and for doing the advanced things offered by, say, VMWare's Dynamic Resource Scheduler are more ambiguous.
Question for the community: Are you moving to "full data center virtualization"? What does that question mean to you? How far are you on getting there and what's standing in your way?
Data Center Virtualization is/will have a profound impact on your IT operations. Everything from how hardware will be procured (who specifies, who owns the budget) to how security threats will get managed to how software licensing is tracked (and charged) to how the application and OS lifecycle gets managed has to be rethought.
Question for the community: Have you started working through this change? How long do you think it will take?
Data Center Virtualization also forces a rethinking of the architecture of the data center at a profound level. It promises to eventually create a level of simplicity, automation and efficiency never before seen in enterprise data center. But, how do you rearchitect your data center to achieve this nirvana? Do centralized SANs still make sense? How does my networking model have to change given the automated migration of applications? How do I know where my assets are and how they're being used? How do I assure sufficient protection of critical data given this new approach? Can I greatly simplify my disaster recovery/business continuity solution using virtualization?
Question for the community: What are the biggest opportunities for you to gain from the virtualization transition in the long term? What are your biggest challenges?
Every new technology has a thousand little gotchas that nobody thought of before it started. Virtualization gives great keynote ...
Question for the community: What are the little things that you're running into that make your life miserable.
As I hope is obvious, I need your feedback. I commit to use this forum on-going to work these issues through and to let you know what role Intel will play in the solutions.