Improving High Throughput Computing’s Energy Efficiency - a Measurement-Based Case Study

Damu Ding will be presenting our research as part of the CQUANDRI project at LF Energy!

Data centers play a key role in cloud computing, and their growing energy consumption leads to significant carbon emissions. To reduce carbon emissions, it is vital to understand their power consumption trends and find effective solutions to energy saving without any significant changes in existing IT facilities, and while still providing sufficient computing performance. This presentation introduces a case study of improving the energy efficiency of high throughput computing, based on power measurements from a UKRI STFC Rutherford Appleton Laboratory (RAL) data center, and focused power optimization measurements at Oxford University. Our study provides an insight into some of the challenges faced by data center operators when trying to reduce carbon emissions. It further discusses how tuning server configurations affects processing job’s energy consumption. We provide practical recommendations for optimizing servers for energy reduction, with a combined view of both performance and carbon emissions. This is joint work of Oxford University, STFC, Bristol University and Newcastle University.