Interview: A Closer Look at Server Efficiency with Klaus-Dieter Lange

Server efficiency means a lot to those who have seen the energy bill of a server room or a data center. Running 24/7 and often at top capacity, servers draw a lot of electricity. However, things are not as dark as they may seem. There are ways to monitor and improve the energy efficiency of your server room.

The Server Efficiency Rating Tool (SERT) created by SPEC is designed to measure server energy efficiency for a wide range of configurations. SERT uses a set of synthetic worklets to test discrete system components such as processors, memory and storage, providing detailed power consumption data at different load levels. It is an integral part of the US Environmental Protection Agency (EPA) ENERGY STAR for Computer Servers program.

Our Luminext team interviewed Klaus-Dieter Lange, chair of the SPECpower committee that develops SERT, about server technology, server use and the SERT tool itself.

Q1.) There is no question that SERT is a remarkable tool. With the EPA supporting SPEC and the collaboration of tech leaders such as AMD, Dell, Fujitsu, HPЕ, IBM, Intel, Microsoft, and Oracle, this tool probably covers it all. Are these companies satisfied with the way SERT benchmarks the performance of their products?

Thank you for your kind words. I cannot speak for the individual companies that for many years have contributed engineering resources for the development of SERT. But speaking as chair of the SPECpower committee, this widespread company participation ensured that we created a tool that challenges every company equally to achieve higher server efficiency, regardless of OS, platform or server configuration. Another indication of SERT fairness is the adoption by the U.S EPA, and the consideration for adoption by the European Commission's Lot9 initiative, as well as the China National Institute for Standardization (CNIS).

Q2.) SERT benchmarks the default server configuration. However, using a server sometimes involves both hardware and software tweaks. Will a future version of SERT ever account for a modified server configuration? ­Especially if the modifications are supposed to be permanent.

The requirement to measure server efficiency on a default server configuration is stipulated by the U.S. EPA ENERGY STAR for Server program in order to encourage server manufacturers to set default server settings to a high-efficiency mode. The SERT is designed and tested across many platforms, OSs and configurations, and accommodates modified server configurations for those doing testing outside the EPA guidelines.

Q3.) Treating the server as just a machine seems hardly fair. Would you agree that in many cases, server efficiency depends a great deal on the people running it? Can responsible use lead to greater energy saving?

Yes, of course, the people running the servers play a role, but SERT can give people valuable guidance on how to configure a workload and utilize servers for better efficiency. The SERT workload covers a wide array of worklets that stress individual subsystems of the server at multiple load levels.

Q4.) How fast are data centers in making improvements and upgrades? ­Servers are neither cheap nor easy to come by, so how much time does it usually take to see a report with improved stats?

My experience shows that modern data centers utilize multiple generations of servers, upgrading the old generations every 1-2 years.  The SERT output data can guide users in their purchasing decisions. Many times, the price for the new servers is compensated for by the reduced cost of their energy usage after several months.

Q5.) At what scale is energy­ efficiency most problematic? Are large data centers the ones with a higher percentage of energy leaks or small server rooms?

That really depends on the resources that are available and the oversight of the facility. A smaller server room might have only one administrator, for example, where a larger data center most likely has a team of engineers available. Both scenarios have their own challenges, and often the large data center has greater resources and a wider range of solutions available to engineers who are overseeing it.

Read More Interviews:

› Server Efficiency