A Combined Dual-stage Framework for Robust Scheduling of Scientific Applications in Heterogeneous Environments with Uncertain Availability
Open Access
- 1 May 2012
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 21, 193-207
Abstract
Scheduling parallel applications on existing or emerging computing platforms is challenging, and, among other attributes, must be efficient and robust. A dual-stage framework is proposed in this paper to evaluate the robustness of efficient resource allocation and dynamic load balancing of scientific applications in heterogeneous computing environments with uncertain availability. The first stage employs robust resource allocation heuristics, while the second stage incorporates robust dynamic loop scheduling techniques. The combined dual-stage framework constitutes a comprehensive framework that enables and provides guarantees for the robust execution of scientific applications in computing systems where uncertainty is caused by various unpredictable perturbations. The paper reports on studies for determining the best techniques to be used for each stage that: (a) maximize the probability that the system make span satisfies a deadline, and (b) minimize the system make span for every given availability level in the system. The usefulness and benefits of the proposed framework are demonstrated via a small scale example.Keywords
This publication has 12 references indexed in Scilit:
- Investigating the robustness of adaptive Dynamic Loop Scheduling on heterogeneous computing systems2010
- Stochastic-Based Robust Dynamic Resource Allocation in a Heterogeneous Computing System2009
- Dynamic load balancing with adaptive factoring methods in scientific applicationsThe Journal of Supercomputing, 2007
- Grid Resource Allocation and Task Scheduling for Resource Intensive Applications2006
- Measuring the robustness of a resource allocationIEEE Transactions on Parallel and Distributed Systems, 2004
- Determining the Execution Time Distribution for a Data Parallel Program in a Heterogeneous Computing EnvironmentJournal of Parallel and Distributed Computing, 1997
- Parallelization of DOALL and DOACROSS Loops—a Survey1997
- Factoring: a method for scheduling parallel loopsCommunications of the ACM, 1992
- Allocating modules to processors in a distributed systemIEEE Transactions on Software Engineering, 1989
- Heuristic Algorithms for Scheduling Independent Tasks on Nonidentical ProcessorsJournal of the ACM, 1977