Implementing green technologies and practices in a high performance computing center

In this paper we describe a comprehensive strategy of modelling, sensing, and using energy-aware technologies to improve the operational efficiency of a high-performance computing (HPC) data center. Modelling was performed using state of the art computational fluid dynamics (CFD) for optimizing air flow and air handler placement. Also included were the installation of a complete environmental monitoring system to provide the real-time tools needed to monitor operating conditions throughout the data center and utilize this information to calibrate the CFD models and realize energy savings. This system was used to implement subfloor leak management, isolation of unneeded regions, perforated tile optimization, and bypass airflow management to maximize the efficiency of the center's cooling infrastructure without adding additional power. In addition, given the high-density of servers in a typical HPC data center rack, rack-based chillers that more efficiently deliver cooling to the servers were installed, allowing several of the large data center air handling units to be shut down. These changes, along with implementing a policy of energy awareness in major and minor equipment purchases, allowed us to achieve a remarkable ten-fold increase in computational capacity while simultaneously reducing the total energy consumption by more than 20%. The strategies and techniques utilized here are readily transferable to other existing data centers.