I got asked to do a 40 minute slot on Virtualisation and cloud at a partner event to day run by EZEIT, while Simon got the tough job of condensing windows deployment down to  40 minutes as well (his post and deck is here)..

I got a bit flat footed here, as the audience were already familiar with Hyper-V, and live migration which is what I had planned to demo, and actually the demos were in the Reading Microsoft Technology Centre and worked flawlessly.  I simply maxed out a SQL Server VM with a simply repetitive query and showed this in perfmon and then live migrated this in System Center Virtual Machine Manager.

Fortunately the rest of my talk was on public and private cloud as you can see from my deck.  The public cloud message is getting out there, although each vendor has there take on it and I’ll come back to that in several more posts. It’s the private cloud that causes most confusion and if I am honest not all of my colleagues get this, and I certainly didn’t until I had coffee with Zane Adam the general manager for Azure.

He simply makes the point is that it’s the next step on the road following server virtualisation, and the dynamic data centre.  However it’s not that straightforward to grow your own although there many of the essentials components in the latest versions of the System Center Suite:

  • SCVMM now has a self service portal to allow users to provision VMs and be charged monthly for them.  It also has an understanding of workloads and knows the best location for a new VM.
  • Service Manager can create tickets for errors and issues that arise in System Center Operations Manager as well as user issues.
  • Opalis can create workflow for management tasks.

However as at today (November 2010) there is no actual private cloud software you can buy, and although the Azure appliance is coming this is only suitable for the biggest business at the moment.  Actually this concept is difficult to achieve anyway because the  IT manager/CIO/CTO is back in the business of trying to buy enough computing power to meet predicted needs. Imagine if a hypothetical System Center Private Cloud product did exist which could provision extra services on demand you would still actually need idle hardware resources to run the extra services and surely no one deliberately has this spare capacity lying idle, unless by a fortunate coincidence another service has been throttled back at the same time.

That’s not to say that the service model of computing can’t be applied today for example charging business per user per month for access to on premises services or based on data volumes, just that the automatic scalability and elasticity associated with public cloud is difficult to achieve in all but the largest businesses.