More Bang for Your Buck: Boosting Performance with Capped Power Consumption
Juan Chen,Xinxin Qi,Feihao Wu,Jianbin Fang,Yong Dong,Yuan Yuan,Zheng Wang,Keqin Li
Table 1 Parameters used by our power-performance model.
 Notation Description $c$ Number of processor cores on each computing node $n$ Number of processes $N$ Number of assigned computing nodes with the default resource allocation strategy $Δ?N$ Number of computing nodes to be increased $bi?(t)$ Memory traffic of computing node $i$ at time $t$ $BN?(t)$ Average value of $bi?(t)$ for $N$ nodes, $0⩽i⩽N$ $BN$ Maximum value of $BN?(t)$, $0⩽t⩽T$ where $T$ represents program running time on $N$ nodes $B$ Physical memory bandwidth on a single computing node $α$ Threshold for the ratio of memory traffic to memory bandwidth $fmax$ Maximum CPU clock frequency $fmin$ Minimum CPU clock frequency $fi$ Processor clock frequency, which satisfies $fmin⩽fi⩽fmax$ $Pcpu?(fi)$ Power consumption of one CPU core with clock frequency $fi$ $Pidlecpu$ Power consumption for each CPU core in the idle state, which equals $k⋅Pcpu?(fmax)$ $Pmem$ Power consumption of DRAM on one computing node $Pother$ Power consumption of compoents other than the CPU and DRAM on one computing node