Sunday, 29 June 2008

Hyper-V on Every Box in the Enterprise

Yes, it's more Sunday evening than afternoon but I'm gonna squeeze this one in under the guise of another installment of Sunday Afternoon Architecture And Philosophy.

With the release of Hyper-V 1.0, I got to thinking... why wouldn't you put it on every x64 physical box in the enterprise? Theoretically, the parent partition should incur no virtualisation overhead - it has physical hardware access and doesn't need to worry about managing VM's if it's running on it's own. It doesn't cost anything more for software assurance customers, and let's face it - if you're running a physical box today (and according to IDC and the like, 80-90% of servers are still physical) then almost all the reasons for choosing ESX over Hyper-V simply don't apply as you don't get any of those features with physical boxes anyway. So if that phsyical box is running at 5% resource utilisation and the owner has a need for another environment (be it for dev, UAT, tiering an app, whatever), if we have Hyper-V installed already then we have the potential to exceed expectation by quickly provisioning a box for them that doesn't require the (usually) lengthy process of ordering and racking. If we think of virtualisation purely in terms of maximising hardware utilisation, then as far as I'm concerned there's no argument to be had - I'd rather have one less physical box by running it on Hyper-V than not having Hyper-V at all.

Of course we all know there are so many other benefits to be had by virtualising on VI3, but as I said if only 10% of boxes are virtual currently then we're missing something. Maybe your business is still uncomfortable with virtualisation. Maybe your chargeback model is so close to the cost of physical boxes that your business doesn't care to virtualise (naively ignoring the cost of rack space, power, cooling and having to build a new datacenter when you run out of those). I'm sure there are a myriad of reasons, and even more certain that none of them are technical. But if we can leverage Hyper-V by putting it on every box and giving our business customers a taste of virtualisation, then maybe we can effect a change in attitude. And maybe next time they go to buy that physical box, they'll ask around as to how that virtual box has been running and reconsider. We've tried the 'boil the ocean' approach with virtualisation over the past few years, and it clearly hasn't worked if only 10% of servers are virtual. It's time for baby steps.

So to wrap it up, let me be absolutely clear - I'm not suggesting we start migrating our virtual infrastructures to Hyper-V now or in the forseeable future. What I am suggesting is that we don't write off Hyper-V by comparing it with VI3 and closing the door on it immediately. A slightly unconventional use case it may be, but it could be the most valuable tool we have had so far in assisting with the virtualisation of that other 80% of the enterprise. And I certainly hope that VMware will imminently release an update to Converter to provide full support for V2V from Hyper-V ;-)