ICT managers’ efforts to hold down escalating server power and cooling costs are being stymied because hardware vendors report energy efficiency in different ways, making it difficult to compare products.
However, that’s set to change. The US EPA (Environmental Protection Agency) and a group of major vendors are planning to release a final draft of a protocol for measuring the energy efficiency of servers. The protocol could be formally approved and in use within a year, but the leader of the effort to develop it says its users will need to push their vendors to adopt the standard.
“It’s going to take some pressure by the customers. If the customers demand it, the manufacturers will change really fast,” says Jonathan Koomey, a consulting professor at Stanford University and a staff scientist at Lawrence Berkeley National Laboratory. He has advised the EPA on energy issues and was tapped by the agency to head the server protocol effort.
One large customer who is already asking for standardised energy-efficiency data from vendors is Lehman Brothers.
“From an end-user standpoint, a standardised method of establishing and reporting server-energy consumption is essential,” says Rick Salmon, vice president of the mission-critical facilities group at the New York-based financial services firm.
A common energy-efficiency protocol should also promote vendor competition and spur efficiency innovations, Salmon says, adding that even a small reduction in power utilisation by servers can have a significant impact on datacentre costs.
Rackspace, a managed hosting services provider in San Antonio, operates more than 20,000 servers. Paul Froutan, the company’s vice president of research and development, conducts his own tests on server energy use and estimates that there may be as much as a 30% variation among similar systems.
But, like Salmon, Froutan says even slight improvements in energy efficiency could be a boon for corporate electricity budgets.
For instance, if each of Rackspace’s servers used about 200 watts of energy per hour and that consumption level was reduced by five watts, or 2.5%, that would save 100 kilowatts per hour — enough to power 500 servers, he says.
Bill Leo, CIO at Mercer Delta Consulting, says he has doubts about the willingness of server vendors to challenge one another over energy efficiency in their sales efforts. “Unless they are going to be able to really compete with each other in this area, I don’t know if [the protocol] is going to have a whole lot of value,” Leo says.
In addition to the EPA, server vendors such as IBM, Dell, Hewlett-Packard and Sun Microsystems are involved in the effort to develop the protocol, which began earlier this year.
Many ICT managers aren’t sure what they spend on electricity because that isn’t part of their budgets. Instead, it may be counted within overall facilities budgets.
As a result, executives such as Dick Hamann, CIO at Seminole Community College in Florida, are working to both reduce power costs and get a better understanding of the actual usage of electricity within their departments.
Based on one of the recommendations made by a consultant who studied Seminole’s power usage, Hamann has, over the past six months, been gradually equipping the school’s PCs with automated shutdown software developed by Persystent.
He plans to have the software installed on 3,500 systems by the end of the year. According to the study, the technology may save Seminole as much as US$65,000 (NZ$97,000) in electricity costs per year, he says.
Scott Townsend, CIO at Michigan Schools and Government Credit Union, says that when buying servers, he focuses on factors such as memory and speed, not energy costs. But if server makers adopt an energy efficiency standard that enables ICT managers to make useful comparisons of systems, “it will definitely have an impact on our purchasing patterns”, Townsend says.