Greening the data centre (for fun and profit)
There’s a lot to be said for the modern data centre movement, and the trend away from power-hungry inefficient racks of under-utilised servers. Virtualisation lets you pack more and more functionality onto each CPU, saving space and power (at the expense of a little more cooling, or a slightly shorter server lifespan). It’s one of the themes here at the Microsoft Management Summit, where the Redmond giant is showing off its next generation virtualisation tools.
With a conference centre full of IT professionals from all over the world, there’s an understandably sceptical audience. After all, a lot of the tools that Microsoft has announced will change the way they work – and how they relate to the rest of the business. It’s a sea change in the way IT is delivered, and one that requires a considerable amount of maturity in the way businesses deliver IT. Instead of applications, we’re finally going to have to live up to the service mantra the industry has been chanting since the turn of the century, designing to contract and designing for failure.
There are plenty of benefits to this approach. For one thing it lets us design for scale out architectures, able to handle system degradation and to support rapid changes in business demands. It also lets us move to self-service delivery and a closer relationship with the owners of the business processes that we’re supporting (and it also changes and simplifies our security model from server-centric to information-centric, making it much easier to meet the demands of regulatory compliance).
It’s still a big change, and we’re going to need proof points to help us get started – proof that there’s gold in them thar hills.
So it’s a good thing that Microsoft has put its money where its mouth is with the infrastructure it’s using to drive its management tools event.
We spent some talking to Jeff Woolsey from the Hyper-V team about the systems that are delivering the conference’s hands-on labs. Just a couple of years ago they filled several racks, took multiple trucks to deliver from Redmond, and needed additional power and cooling. Last year, by switching to a heavily virtualised environment, they needed just two racks, with separate VM clusters for each function. There wasn’t any need for extra power, and the racks could be cooled using just the standard hotel air conditioning – letting Microsoft display the machines in a “severquarium”, right next to the machines that delivered the hands-on labs.
2011 has seen a further evolution of the servers. Still using two racks, they’re taking advantage of the latest processor technology to add cores, and reduce the amount of cabling. New management tools mean they can be managed as a single fabric, and Hyper-V R2’s dynamic memory features mean that Microsoft is able to run more labs using 1TB less memory than in 2010. At current memory prices that reduction results in a significant saving – around $800,000.
Saving money is good, saving space is better. Virtualisation and the shift to thinking about your infrastructure as a fabric hosting private clouds is a route to delivering both. You get to be green AND you get to both save and make money, a win in anyone’s book. With a new generation of tools that simplify building and deploying virtual machine hosted services, it’s time to finally stop thinking about changing the way we do IT and actually go out and do it.
Simon Bisson