The storage industry and the U.S. Environmental Protection Agency are crafting energy consumption standards for enterprise storage as that part of the data center becomes a bigger target for questions about efficiency.
On Thursday, the EPA will hold a public meeting here at Storage Networking World in Orlando to solicit input on its developing Energy Star Data Center Storage specification. It released Draft 1.0 of the standard on April 9 and plans to conduct more meetings to refine that document, which may be finished by the end of the year. Storage vendors will then be able to label their products as compliant with the energy-saving standard.
Servers were the first focus of the green movement in data centers because they traditionally have consumed the most power of all components there. But through virtualization, Energy Star certification and utility-sponsored incentives to reduce the number of servers in data centers, the computing side in many cases has slashed its energy usage. In some cases, that has left storage systems eating a bigger piece of the energy pie in given facilities, said S.W. Worth, a senior standard program manager at Microsoft, who also works with the Storage Networking Industry Association (SNIA). As a result, they could come under greater scrutiny from business managers trying to reduce costs, he told a breakout session at SNW on Wednesday.
Setting energy standards for storage systems will require a whole effort of its own because they work differently and serve a different purpose from servers, said Erik Riedel, a member of the SNIA Technical Council and a senior director of technology and architecture at EMC.
Whereas servers are essentially just tools to carry out various computing tasks, each storage platform is the repository for specific data, Riedel said. This has implications for how they work. For example, storage systems almost always have dual power supplies, whereas servers often don't. If the sole power supply on a server were to fail, its work could be shifted to another server. But if the same happened on a storage device, applications would lose access to the data on that platform until it came back up, Riedel said during a session at SNW.
In addition, the idle state on a storage device is different from that of a server. For example, idle hard drives carry out housekeeping processes such as going through the data to correct errors and rearranging blocks of data for easier access in the future, Worth explained. This is one reason why storage power consumption at idle is usually less than 20 percent lower than in the active state, he said.
It has taken the EPA some time to learn about these differences, according to Riedel. Now the agency is mapping out a complex taxonomy of different types of storage systems based on size, performance and design, and developing different energy consumption requirements for Energy Star certification for each type, Riedel said. At Thursday's meeting, it will be looking for feedback from storage professionals on how it has crafted those classifications. Comments on Draft 1 of the standard are due to EPA by May 21.
The EPA has been holding public meetings about once every quarter, and though the length of the standards process is up to the agency, it's possible vendors will be able to label their products as Energy Star compliant late this year, Riedel said.
There are many areas where vendors can cut the energy consumption of their systems, Riedel said. Power supplies, disk arrays, hard drives, controllers and other components all could be made more efficient. But the biggest breakthroughs are likely to come from higher up in the system, he said. For example, if the storage system knew that the virtual machine associated with a particular application isn't being used, then the disks that hold the data for that application could be spun down until the virtual machine came on again, he said.
"It's that feedback loop, and that total stack, that's going to give you the biggest bang," Riedel said.