UC San Diego

This thesis presents Green Queue, a production quality tracing and analysis framework for implementing application aware Dynamic Voltage-Frequency Scaling (DVFS) for MPI applications in high performance computing. Green Queue makes use of both intertask and intratask DVFS techniques. The intertask technique targets applications where workload is imbalanced, reducing frequency for ranks with smaller workloads. The intratask technique targets balanced workloads where all tasks are synchronously running the same code. The strategy identies program phases and selects the energy-optimal frequency for each by predicting power and measuring performance response to frequency changes. Green Queue is evaluated at 1024 cores on Gordon, the newest supercomputer at the San Diego Supercomputer Center, utilizing Intel Sandybridge CPUs. Green Queue demonstrates up to 21% and 32% energy savings for intratask and intertask DVFS strategies, respectively