SC13 Home > SC13 Schedule > SC13 Presentation - P-SLURM: Power-Aware Job Scheduling on a Hardware-Overprovisioned HPC Cluster

SCHEDULE: NOV 16-22, 2013

When viewing the Technical Program schedule, on the far righthand side is a column labeled "PLANNER." Use this planner to build your own schedule. Once you select an event and want to add it to your personal schedule, just click on the calendar icon of your choice (outlook calendar, ical calendar or google calendar) and that event will be stored there. As you select events in this manner, you will have your own schedule to guide you through the week.

P-SLURM: Power-Aware Job Scheduling on a Hardware-Overprovisioned HPC Cluster

SESSION: Doctoral Showcase - Early Research Showcase

EVENT TYPE: Doctoral Showcase - Early Research Showcases

TIME: 3:30PM - 5:00PM

SESSION CHAIR: Wojtek James Goscinski

Presenter(s):Tapasya Patki

ROOM:601/603

ABSTRACT:
While traditional supercomputers are worst-case provisioned and guarantee peak power to every node, next-generation supercomputers will require hardware-overprovisioning and intelligent power scheduling. Our results demonstrate that choosing the optimal configuration (number of nodes, number of cores per node, power per node) based on individual application characteristics on an overprovisioned cluster can improve performance under a power bound by up to 62% when compared to worst-case provisioning. In this work, we propose P-SLURM, a power-aware resource manager for an overprovisioned cluster. P-SLURM takes job-level power bounds into account to implement policies that choose job configurations based on metrics such as average turnaround time and utilization. In addition, it facilitates reallocation of power at runtime as jobs enter and leave the system. We extend SLURM's easy backfilling algorithm with seven power-aware policies, and present some early results obtained by using job traces gathered from Lawrence Livermore National Laboratory.

Chair/Presenter Details:

Wojtek James Goscinski (Chair) - Monash University

Tapasya Patki - University of Arizona

Add to iCal  Click here to download .ics calendar file

Add to Outlook  Click here to download .vcs calendar file

Add to Google Calendarss  Click here to add event to your Google Calendar