Optimize data center energy use: improve cooling, processor and server efficiencies to reduce power consumption

Communications News, August, 2008 by John Schmidt

Data center energy use has become a mainstream issue for manufacturers, data center managers and even legislators. Data centers accounted for 1.5 percent of all electricity use in 2005, according to the EPA, and according to IDC, in 2010 there will be an estimated 15.8 million servers in the United States, triple the number that existed in 2000.

The impact of this growth has made the importance of optimizing the energy used in data centers a primary issue for manufacturers providing everything from servers to structured cabling. As such, understanding some of the new methods that can be used to optimize data center efficiency is important.

Gartner estimates that only about 1 percent of the energy generated to power a data center ever makes it to the processor. Power loss in the electrical-distribution network, power-conversion losses and cooling reduce the percentage of power that makes it into the processors doing the computational work of a data center.

More efficient processors reduce the heat generated relative to the processing power. Advances such as 45-nanometer silicon and quad-core processors can reduce the power consumed by the processor. Intel cites a 44 percent reduction in power costs when using quad-core technology compared with dual-core processors.

Server virtualization is another area of innovation that can positively impact energy optimization. Server virtualization allows multiple applications to run simultaneously on a single machine, allowing consolidation and reduction of servers. Idle servers still consume a significant amount of power, so maximizing the use of each server will result in overall power reduction.

VMware, a maker of virtualization software, provides an online calculator that indicates virtualizing 30 servers can reduce the energy costs for a data center by $20,000 per year, and will reduce C[O.sub.2] emissions by 266,282 pounds a year.

Approximately 50 percent of the energy used within a data center is consumed by cooling equipment. Technologies that are more efficient at removing heat from the server and server cabinets will have a positive impact on the power efficiency of the data center.

A logical first step is to have a computational fluid dynamics analysis performed on the data center. This computer modeling can predict areas of poor airflow, increased heat and wasted cooling.

Technology advances also exist in the area of computer room air conditioning itself. Traditional methods of cooling consist of flooding a raised floor with chilled air from computer room air conditioner units. This method will tend to cool the entire computer room, resulting in wasted energy.

Best practices, such as hot aisle/cold aisle placement and hot aisle containment, will improve efficiencies. New technologies drive cooling closer to the intended point of use, the server. In-row cooling is a concept that puts the air conditioning units adjacent to the server cabinet. This can reduce annual electrical costs for air conditioning by 50 percent or more for average rack power densities greater than 6 kW when compared with room-oriented cooling.

Poor cabling management can restrict airflow by blocking air inlets and exhausts. Excess cables that are not removed and excess lengths of cabling that are not properly managed contribute to airflow restriction. Paying close attention to cleaning up these areas, specifically around active equipment, will improve airflow and will also make the installation easier to manage. Removing raceway systems from the raised floor and positioning them overhead removes potential air dams from underneath the floor, thus improving the airflow.

There is no single solution that will solve these energy efficiency issues. Combined efforts with regards to processor and server efficiency, cooling and optimized airflow to reduce the power usage will be necessary.

John Schmidt is senior product manager of business development for TrueNet Structured Cabling Systems, ADC, Minneapolis, Minn. Schmidt holds an engineering degree from the University of Minnesota and has 10 issued patents for telecommunications and network equipment design.