The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project will result from development of a thermal management technology that could save 20% of electricity consumption and increase equipment life-time by 25% in data centers. In 2013, U.S. data centers consumed approximately 91 billion kWh of electricity and caused 97 MMT of carbon pollution. By 2020, the data center annual electricity consumption is projected to increase roughly to 140 billion kWh and cause 150 MMT carbon pollution. Inefficient operational energy usage and over-provisioned capacity in data centers theoretically offer the potential for 40% energy savings. The proposed technology improves data centers' energy efficiency and capacity utilization and realizes half of the potential savings. A major barrier to realizing this savings potential is the risk aversion toward data center downtime that costs $7, 900/min, with 91% data centers experiencing an average 86 min outage per year. The proposed technology ensures reliability by safeguarding data centers from downtime due to overheating. This SBIR Phase I project is poised to have a far-reaching societal impact by implementing green methods in the ubiquitous information technology and related service industries.
This Small Business Innovation Research Phase I project will demonstrate the feasibility of a software appliance (SA) that addresses cooling optimization for data centers. Due to time-scale differences between IT and cooling operations and diversity of interrelated equipment, data center operators fail to determine reliable cooling set-points in real-time and are compelled to operate with low IT utilization and over-provisioned cooling. The company's software appliance determines the most cost-efficient cooling set-points, that ensure reliable computing operations, in real-time and automatically implements them through the building management system. Given that the proposed SA promotes workload-proportional cooling, it avoids wasteful cooling resource peak provisioning. The development of the proposed SA will involve the following activities:
- development of a real-time predictive analytics framework;
- extraction of critical data (e.g. IT device temperature data) for different workload types and cooling set-points; and
- transferring optimal cooling set-points to the building management system to control cooling systems.
The anticipated result for the project is an SA demonstrating proof-of-concept, and feasibility of the proposed technology in cooling energy usage reduction and cooling capacity utilization improvement of data centers.
