Operation-Aware Power Capping.

Once the peak power draw of a large-scale high-performance-computing (HPC) cluster exceeds the capacity of its surrounding infrastructures, the cluster's power consumption needs to be capped to avoid hardware damage. However, power capping often causes a computational performance loss because the underlying processors are clocked down. In this work, we developed an operation-aware management strategy, called OAPM, to mitigate the performance loss. OAPM manages performance under a power cap dynamically at runtime by modifying the core and uncore clock rate. Using this approach, the limited power budget can be shifted effectively and optimally among components within a processor. The components with high computational activities are powered up while the others are throttled. The overall execution performance is improved. Employing the OAPM on diverse HPC benchmarks and real-world applications, we observed that the hardware settings adjusted by OAPM have near-optimal results compared to the optimal setting of a static approach. The achieved speedup in our work amounts to up to 6.3%.