Virtualisation and Business Value
Several big virtualisation vendors (or their business parters) are contacting me to sell their wares. They have come from just about every angle, Green IT, cost, ease of use, scalability, supportability and the list goes on. Recently one vendor invited me to use their own online calculator to estimate the cost savings using their virtualisation platform. After entering a myriad of details it spat back an answer showing a -32% cost saving for our current non-virtualised infrastructure. That’s right, an increase in cost of 32% to adopt virtualisation. It’s either an error or they value honesty. Note their sales guy was in earlier extolling the benefits of virtualisation across the business though. The reality is, if I choose to virtualise any part of my infrastructue with these vendors I am up for a cost. I have consulted to several companies recently where the executive management has been convinced that virtualisation will save signficant amounts of money on their current infrastructure and their current IT costs will immediately drop. Nothing could be further from the truth. Whether it is subscription/license fees, consulting, or the significant time required internally on deploying and reconfiguring systems, the full cost will be measured in tens of thousands of dollars. In this current economic environment I would perhaps choose to spend that money on revenue earning opportunities. If you are at the initial build or update of your systems, virtualisation may well have cost savings over a period but not immediately.
Something for nothing
Many virtualisation sales consultants have used the “something for nothing” pitch, that generally all servers are under utilised and CPU usage should be at least 80% with at least two systems working on a single box. Well 80% doesn’t give me much wiggle room for peak loads but something else bothers me. If you have ever worked on a machine where the CPU usage is a sustained 80% or so the performance drop is very noticeable and sometimes unusable. Depending on the application before we get to 80% CPU usage we will probably hit an i/o limit and impact all virtual instances on the box. Hardly something for nothing. With this in mind I would have a significant effort on my hands to stage, test and tune production applications before implementing a virtualised approach. The work and cost is adding up. What happens if my two virtualised instances unexpectedly hit peak demand together? I know of one instance where this occurred and the virtualisation vendor almost flippantly suggested to purchase a new server.
I’m not buying a drill, I’m buying a hole
As a business, I don’t want virtualisation, I want the benefits it is supposed to deliver. I need to consider those benefits in the full context of the business and the financial position of the company given the economic times. You will probably find that unless you are starting from scratch with your production servers or at an upgrade point it might be wiser to use the money on a revenue-earning investment elsewhere in the business.
Anyone with thoughts or experiences on this topic, I’d love to hear from you.
