Energy efficiency in high-performance computing (HPC) data centers needs to be viewed from the thousand-foot perspective, according to scientists at the Leibniz Supercomputing Centre of the Bavarian Academy of Sciences and Humanities in Munich, Germany.
Typical energy management at HPC data centers does not permit data sharing between the HPC system, the hosting facility and infrastructure, according to an article in HPC Wire. The scientists say there are no tools that “support a complete collection and correlation of energy efficiency relevant data, allowing for a unified view of energy consumption present in the datacenter,” reports HPC Wire.
To address this concern, researchers at the Leibniz Supercomputing Centre are developing a new energy measuring and evaluation product – Power Data Aggregation Monitor (PowerDAM) – to monitor and correlate data from the HPC systems as well as cooling and building infrastructure. PowerDAM will monitor both physical sensors as well as “virtual sensors” and provide visualization for factors such as power draw, utilization rate, and average CPU temperatures. In addition, PowerDAM will report on datacenter key performance indicators such as PUE, ERE, DCiE and WUE, reports HPC Wire.
In the US, some leading-edge HPC data center work is being conducted at NREL in Golden, Colo., where the on-site HPC data center is using warm-water liquid cooling to reach an annualized average power usage effectiveness (PUE) rating of 1.06 or better. NREL’s data center also captures and re-uses the heat generated by the computers in the HPC system for the primary heat source of some office space and laboratories.
Photo: Data Center via Shutterstock