In a conversation last night with John Swainson, CEO of CA, I asked a question about optimization. Here’s the context:

CA is making a big noise about the optimization of virtual machine deployment. I had been asked by a journalist whether CA wasn’t a bit late to the virtualization party, given that the IT press has been awash with virtualization for several years and a whole bevvy of IT start-ups have materialized to virtualize everything from the desktop to the watercooler. In fact, for the last few years, the advice to start-ups from the VC community seems to have been:

If it moves virtualize it.

Similarly, data centers across the world have been consolidating servers at such a rate that server farms have been transformed into server small-holdings. Nevertheless, all of that has happened with very little resource optimization – in fact none at all.

The point is that there’s a big difference between:

1. Using virtual machines to increase the efficiency of resource utilization, and
2. Organizing the dynamic deployment of virtual machines in order to optimize resource utilization in an end-to-end manner.

Getting quick wins with VM deployment is relatively easy, especially if you preserve the application silos that already exist so that they become virtual silos. And that is what most organizations have been doing. Indeed some of them have VM server consolidation projects that have been running for quite a while and will run for a year or two yet. If you have thousands of servers, many of them running diverse applications, it can take years to consolidate them all.

So the question I asked John Swainson (who, incidentally, is a highly competent technician) was “what you will gain from optimization if you’ve already consolidated?” Now John, it has to be said, has an axe to grind in that CA has built software to discover application dependencies and deploy VMs dynamically. (I’ll provide more details on this at some point). However he didn’t respond by attempting to sell me on CA’s technology or its efficacy, he gave a more useful answer than that.

He said that, professionally, he’d started out in process control and he spoke from his experience of working in that area. He said that when you truly optimize a complex end-to-end process, somewhere between 15% and 20% cost reduction is possible and consequently he expected that VM-based resource optimization would achieve gains in that area.

Now that’s a nice rule of thumb and until someone disputes it, I’ll use it: Optimization of a process gets you a 15 to 20% win.