An intelligent energy optimization approach for MPI based applications in HPC systems

Energy-aware computing is gaining more and more attention in high performance computing (HPC) environment. As an outcome of this, various energy-aware techniques are existing and many are being proposed. But it is difficult to have a technique which saves energy without compromising the performance. This paper talks about a novel energy optimization approach for Message Passing Interface (MPI) applications running on HPC systems. Our approach relies on applying Dynamic Voltage Frequency Scaling (DVFS) at node level by an optimization agent. Whenever MPI processes are idle or busy with I/O operations, the corresponding CPU cores run at higher frequencies, which results in wastage of power. During this time, CPU cores frequencies can be reduced using DVFS so that the energy can be saved. Our approach is based on a Multi-agent based intelligent energy management framework, which uses an optimization agent for implementing energy optimization algorithm. The key advantage of the proposed approach is that the performance will not be compromised while achieving energy savings.

[1]  Yu Zeng,et al.  Automatic Energy Status Controlling with Dynamic Voltage Scaling in Power-Aware High Performance Computing Cluster , 2011, 2011 12th International Conference on Parallel and Distributed Computing, Applications and Technologies.

[2]  Dong Li,et al.  Hybrid MPI/OpenMP power-aware computing , 2010, 2010 IEEE International Symposium on Parallel & Distributed Processing (IPDPS).

[3]  Yan Ma,et al.  Energy-efficient scheduling algorithm of task dependent graph on DVS-Unable cluster system , 2009, 2009 10th IEEE/ACM International Conference on Grid Computing.

[4]  Hafiz Farooq Ahmad,et al.  Multi-agent systems: overview of a new paradigm for distributed systems , 2002, 7th IEEE International Symposium on High Assurance Systems Engineering, 2002. Proceedings..

[5]  Manish Parashar,et al.  Investigating the potential of application-centric aggressive power management for HPC workloads , 2010, 2010 International Conference on High Performance Computing.

[6]  Vijay Tewari.  Standards for Autonomic Computing , 2006 .

[7]  Dong Li,et al.  Power-aware MPI task aggregation prediction for high-end computing systems , 2010, 2010 IEEE International Symposium on Parallel & Distributed Processing (IPDPS).

[8]  Mahmut T. Kandemir,et al.  Leakage Current: Moore's Law Meets Static Power , 2003, Computer.

[9]  Rong Ge,et al.  High-performance, power-aware distributed computing for scientific applications , 2005, Computer.

[10]  Mateo Valero,et al.  Power-aware load balancing of large scale MPI applications , 2009, 2009 IEEE International Symposium on Parallel & Distributed Processing.

[11]  Gregor von Laszewski,et al.  Towards Energy Aware Scheduling for Precedence Constrained Parallel Tasks in a Cluster with DVFS , 2010, 2010 10th IEEE/ACM International Conference on Cluster, Cloud and Grid Computing.

[12]  Rong Ge,et al.  Performance-constrained Distributed DVS Scheduling for Scientific Applications on Power-aware Clusters , 2005, ACM/IEEE SC 2005 Conference (SC'05).

[13]  Sankararaman Venkatesh,et al.  Implementation of Automated Grid Software Management Tool: A Mobile Agent Based Approach , 2006, IKE.

[14]  Xuejun Yang,et al.  Low Power Optimization for MPI Collective Operations , 2008, 2008 The 9th International Conference for Young Computer Scientists.