Energy Efficiency: Storage Steals A March

For some time now, data centre people have been looking for ways to measure energy efficiency inside the data centre, getting beyond the big figures that simply say how efficient the building is. It looks as if things are finally starting to move – and the storage world seems to be ahead of its server brethren.

It all starts from the need to be more efficient in energy use. The Green Grid proposed PUE (Power Usage Effectiveness) as the measurement of the efficiency of a data centre – sorry if you’ve heard this too many times – dividing the energy going in by the amount that reaches the IT equipment.

That’s been successful – PUE has been adopted in many places, including by the US Environmental Protection Agency (EPA) which has extended its Energy Star scheme to data centres. But PUE is about how efficient the facility is, not about the efficiency of the IT  equipment it holds.

Storage efficiency

Once the power reaches the racks, how do you decide if one setup is more efficient than another, or if one vendor’s proposed architecture uses more power than another’s? There has not been a very effective way of doing this.

One problem is that data centres are all different. One data centre may use more storage, but another may use it more intensively, reading and writing bigger files, or doing it more often.

The energy use in the data centre is divided between different equipment, with around ten percent going to networks, 20 to 45 percent going to storage, and the rest going to the servers. Now server vendors all talk up their green credentials and network vendors have begun to talk green (see these recent announcements from Juniper and Force10).

But it seems like the storage industry may be ahead in providing measurable ways to compare the energy efficiency of one storage system with another. SNIA, the Storage Networking Industry Association has a Green Storage Initiative, which has come up with Emerald, a scheme which proposes to certify the energy efficiency of storage systems.

The EPA has been preparing Energy Star ratings for servers for some time, and announced it in May 2009, but this has been criticised for not going far enough. Another approach to server energy comparison has been to run a standard benchmark such as TPC for transactions, and measure the energy used in achieving it.

Emerald

SNIA’s Emerald seems to take a more methodical approach. It divides possible storage systems into several classes, and has defined benchmark loads for each class. Storage manufacturers can build a system that delivers the required perfrormance under the specified load, and will then be assessed on the energy use of that system.

In the first quarter of 2011, storage systems on sale will begin to have Emerald certifications when they are sold. The assessments and certificates should be reliable, say those involved.

“These results will be audited and customers can use them for comparison,” said Leah Schoeb of VMware, chair of the SNIA Green Storage Initiative. The work has taken three years behind the scenes, she told me, and has involved a lot of debate. “The marketing folks have been waiting for us to roll this out – next year there will be a lot of certifications.”

Industry groups sometimes falter at the point where they deliver their metrics to the waiting world, but Schoeb says this won’t happen with Emerald. Sometimes there are doubts when the industry creates a new benchmark, that vendors and users will optimise to a theoretical target – but IT managers are now too sophisticated to fall into this trap, she believes.

Emerald, it seems, could be a very useful tool.