Wednesday, November 17, 2010

Virtualization is for IT, Cloud is for Applications

We all talk about server virtualization like it has been around forever. The fact that just about every IT vendor has fully embraced server virtualization and has marquee marketing and sales campaigns based on its massive success is hard to ignore. And why not, when you have companies like VMware stating that for every $1 spent on a virtualization license, there is an additional $16 spent in software, hardware, and services? Everyone, of course, wants a slice of the the pie and the great news is that there is no sign of a slowdown.

A quick reality check also shows that the majority of virtualization implementations are still very focused on consolidation that drives reduced capital expenditures. It's a pretty easy calculation for any company to make when they don't have to procure a new server for each IT request that comes through the door. IT operations further embrace server virtualization just for the simple fact that their job becomes easier and the risk of maintaining a rat's nest of individual servers is greatly simplified with server virtualization--a giant win for IT!

But what about the workloads that IT doesn't own and manage? The applications that "run the business," require special attention, and risk revenue or productivity loss if they go offline? Is virtualization the best platform for these workloads? Probably not.

Most IT organization categorize applications based on:

* What they need to operate - These are great server virtualization candidates.
* Investment applications - This is the gray area when it comes to virtualization. If you have already made investments in availability, specialized hardware, recent hardware purchases, etc., these application are most often best left as is.
* Containing stuff that won't virtualize and doesn't change much so folks leave it.
* Killing stuff that you have to dump as otherwise it kills you.

You may have a 35 year old mainframe application or old Cobol code that needs constant change that you have to outsource development since you don't have the expertise and it takes years to change the code and is risky. So when change is required, it gets a new system--maybe Oracle financials or SAP--and the function is migrated, not the code.

This migration process is the ideal on-boarding opportunity for the cloud, which is primarily focused on reducing operational expenses. We can debate the different cloud delivery models for hours (and I don't mean IaaS), but the tipping point is real. Not all workloads are ideal candidates for the virtualization platform. The initial consolidation effort is a no brainer; the tweeners require some careful thinking and will largely depend on the current investment and protection. The contain and kill workloads are ideal candidates that should leapfrog virtualization and move directly to the cloud.