Retrofitting Data Centers for Expanded Capacity and Improved Energy Efficiency

Many existing data centers are operating at peak design capacities and cannot grow any further without incurring substantial infrastructure costs. As an alternative, increased capacity can be gained by improving the efficiency of the existing HVAC and electrical systems. By making the cooling system more efficient and reducing electrical losses, existing power is made available for more servers. Many utilities also have rebate programs that can help finance these upgrades, which continue to save money every year they operate more efficiently.

Integral Group has analyzed hundreds of data centers for specific energy efficiency measures, their impact on energy use, and corresponding peak load reductions. Some of the typical measures include air- and water-side economizing, hot/cold aisle containment, increasing temperature differences, installing variable frequency drives, and high-efficiency uninterruptible power supplies. During this presentation, the speakers will discuss the most commonly implemented measures, typical cost savings, and increased capacity, as well as common barriers to implementation.

Finally, we will present case studies to illustrate some of these approaches and the real-world limitations, opportunities, and savings available to data center operators ready to optimize their existing data center. One of the case studies will include the Earl Warren Hall Data Center at the University of California, Berkeley. The existing data center, cooled by traditional DX Computer Room Air Conditioners with a raised floor and no aisle containment, had exceeded its power and cooling capacities. On design days, the outdoor condensing units (mounted on the roof) overheat and supply air temperatures are not able to be maintained. Higher density racks, at 10 kW per rack located in two rows, experience temperature problems during these conditions, indicating the limits of a traditional raised floor installation. A water-side economizer system has been added with a cooling tower, heat exchangers, pumps and rear door heat exchangers on the most heavily loaded racks, substantially reducing the energy use of the facility. A corresponding reduction in peak load on the CRAC units allowed additional servers to be added to the data center. Efficient operation and data center reliability are not mutually exclusive; in fact, most efficiency upgrades enhance reliability through more effective air distribution, less compressor run time, and, under most operating conditions, more distributed and robust cooling capacity.

Session Presented By: