Knowledge Is Power – And Power Management Needs Data
It’s no good trying to guess your way to a low-energy data centre, says Avocent’s CTO Ben Grimes. You have to instrument your systems and get measurements
When you start looking at reducing data centre power, you think it’s going to involve cool new technology, but the biggest issue is getting good information.
It is. In that environment, if I’m controlling the cheque book, before I buy a technology I need to know which gives me the best return on investment. If I’m making assumptions based on theoretical figures, I’m not doing the best I can for my company. If I have historical data I can lay on the table and analyse. I may be able to say I don’t need to buy some widget because I still have headroom, and spend the money on something else I actually need.
If I can measure my environment, I may not have to build that new data centre in the next six months. I may be able to extend that out and do it further down the road, when the economy turns around.
That’s as big a part of the greening of IT as anything else. If you have real time information you can make good decisions. You can monitor what you have today – and model and design for the future.
Does IT have to get involved with other parts of the business?
Instead of giving IT four walls and saying “Fill it up”, IT and Facilities can work hand in hand to optimise it. We see a convergence of IT and Facilities. They both need models and information to make good decisions.
What does Avocent do about it?
We have a set of power solutions. In the data centre world we have intelligent power distribution units. Power strips you can mount within the rack that allow you to have remote control of each socket, and measure the consumption of power at each socket.
We’re also in servers, measuring power consumption at the device. Avocent delivers the onboard service processor technologies for companies like IBM and Dell, to meet the IPMI (Intelligent Platform Management Interface) standard. If you buy IBM or Dell, we’re probably already in your data centre.
The other part of what we do is around DSView 3. It’s a higher level app that goes out and reads that instrumentation. It stores the information so you can do queries over whatever time base you want.
You can have a historical database on power consumption, grouped at room level, or all the way down to an individual device. You can do predictive analysis, and trending, to tell when you are going to run out of power, or set alarms when you get close to them.
We also have a planning tool, called MergePoint that can model a data centre, to optimise where you put your equipment.
All this can also help with chargeback – if one rack belongs to a business unit, we can tell how many kilowatt-hours it used, and the IT folks can charge that to the business unit.
A power usage dashboard?
Absolutely. The first thing you need to do is instrument. Once you do that, you can measure. Once you can measure you can manage. A lot of people go straight to the management, but if you can’t accurately measure, you can’t accurately manage.