Cloud Computing Expo on Ulitzer

The trigger for this post is a conversation I've had far too often with an IT executive who has  an ambitious plan to leverage hypervisor virtualization to create a new data center infrastructure upon which his entire business would run.  The goals are laudable; dramatic cost reductions, increased availability, decreased time to market (as measured by how long it takes to provision a VM)...all things any sensible business or IT executive wants, right?  But when I asked about their plans for business applications that didn't fit his deployment options (literally small, medium, and large) I got an answer that made me cringe:  "they'll have to".

When did The Business start existing for the sake of IT instead of the other way around?  Did I miss a memo?

Of course it doesn't, and unless your business is providing IT goods or services it never will.  Which brings me around to the title - all the talk about Cloud Computing these days is centered on making an ‘unlimited' number of commodity VM's available quickly and for low cost.  Who among you feels any company in the Fortune 1000 with a reasonable amount of complexity could run their entire business on commodity VM's?  I certainly don't know of any...so it puzzles me that any rational IT executive would think they could make that approach work inside their own data center.

In our experience driving large and complex IT Transformations we have found a several profound Truths:

• 30% of the application portfolio consumes 70% of the data center resources
• If you can make things work well for that 30% you're going to make the business happy
• The remaining 70% of the application portfolio will probably fit just fine into the commodity VM's (and if they don't they probably aren't all that important to the business anyway).

So, looping back to where we started...why drive a data center strategy that only addresses the applications that consume 30% of the infrastructure and doesn't focus on what the business cares about?

Technologies like hypervisor virtualization are just one piece of the puzzle.  The ideal balance requires an approach we call Fit-for-Purpose ... literally making the right resources available to the right application workload at the right time, based on an understanding of business policies and priorities AND the infrastructure resources that are currently available (or underutilized).  It involves leveraging and sharing specialized resources in conjunction with commodity platforms in order to optimize the physics of workload execution.  For example, take a function like XML transformation - very common in business applications, and most frequently performed with software on a commodity compute platform.  Sure, that works fine...but what if you could execute that same function at layer 2 with firmware at wire-speed in an XML accelerator?  My experience shows that you can replace a dozen or more servers with a single appliance and improve performance from 5-10 seconds to 0.5 microseconds.  That's Fit-for-Purpose, and just a single example of a functional optimization technique that can be repeated across different parts of your IT delivery lifecycle.

Truly effective Clouds - internal or external - must apply this principle to enable IT to simplify and reduce cost without impairing the differentiation necessary for business success.