Optimizing job performance under a given power constraint in HPC centers

Abstract

Abstract-Never-ending striving for performance has re sulted in a tremendous increase in power consumption of HPC centers. Power budgeting has become very important from several reasons such as reliability, operating costs and limited power draw due to the existing infrastructure. In this paper we propose a power budget guided job scheduling policy that maximize overall job performance for a given power budget. We have shown that using DVFS under a power constraint performance can be significantly improved as it allows more jobs to run simultaneously leading to shorter wait times. Aggressiveness of frequency scaling applied to a job depends on instantaneous power consumption and on the job's predicted performance. Our policy has been evaluated for four workload traces from systems in production use with up to 4 008 processors. The results show that our policy achieves up to two times better performance compared to power budgeting without DVFS. Moreover it leads to 23% lower CPU energy consumption on average. Furthermore, we have investigated how much job performance and energy efficiency can be improved under our policy and same power budget by an increase in the number of DVFS enabled processors.

Discrete control for ensuring consistency between multiple autonomic managers

“…Some of these researches are based on (i) hardware with voltage and frequency control (e.g., DVFS [11]), (ii) resource allocation: Reducing power consumption by reducing the clock frequency of the processor has been widely studied [5,12], Flautner et al [13] explored a software managed dynamic voltage scaling policy that sets CPU speed on a task basis rather than by time intervals. [14] proposes a power budget guided job scheduling policy that maximizes overall job performance for a given power budget. Many works such as [4,[15][16][17][18] focused on dynamic resource provisioning in response to dynamic workload changes.…”

Linear programming based parallel job scheduling for power constrained systems

“…The budget is given for all system processors. We have shown before that the DVFS technique can improve job wait times in a case of a power constrained system ( [4]). DVFS (Dynamic Voltage Frequency Scaling) is a widely used technique that trades processor performance for lower power consumption.…”

“…With the increase in processor power consumption, the available power has emerged as a new constraint. A policy that assigns CPU frequency to a job selected for execution in such a way that a given power budget is not violated has already been proposed ( [4]). Here we refer to this policy as PB-guided.…”

Cost-Efficient CPU Provisioning for Scientific Workflows on Clouds

“…Task runtimes, when running at a frequency lower than the maximum, can be estimated based on the task's CPU-boundedness [10,8]. Hence, the runtime of a task t at a frequency f is given by:…”

“…where runtime t fmax is the task runtime when running at maximum frequency, f max and the parameter β captures the impact of the CPU frequency on job runtime, ranging between 0 and 1 for jobs with low or high CPU activity, respectively [8].…”