Multi-round scheduling for divisible loads on the heterogeneous cluster systems of multi-core computers

Authors:
Li Xia;Zhong Cheng;Qu Zeng-Yan
Affiliations:
Guangxi University, Nanning, China;Guangxi University, Nanning, China;Guangxi University, Nanning, China
Venue:
Proceedings of the 2nd International Conference on Interaction Sciences: Information Technology, Culture and Human
Year:
2009

Citing 6
Cited 0

UMR: A Multi-Round Algorithm for Scheduling Divisible Workloads

IPDPS '03 Proceedings of the 17th International Symposium on Parallel and Distributed Processing
Multiround Algorithms for Scheduling Divisible Loads

IEEE Transactions on Parallel and Distributed Systems
Scheduling divisible loads in multiple rounds

Scheduling divisible loads in multiple rounds
Optimal Multi-installments Algorithm for Divisible Load Scheduling

HPCASIA '05 Proceedings of the Eighth International Conference on High-Performance Computing in Asia-Pacific Region
Understanding the Impact of Multi-Core Architecture in Cluster Computing: A Case Study with Intel Dual-Core System

CCGRID '07 Proceedings of the Seventh IEEE International Symposium on Cluster Computing and the Grid
Parallel Loop Self-Scheduling for Heterogeneous Cluster Systems with Multi-core Computers

APSCC '08 Proceedings of the 2008 IEEE Asia-Pacific Services Computing Conference

Quantified Score

Hi-index	0.00

Visualization

Abstract

By allowing the overlap execution of computation and communication, a multi-round algorithm with inter-node scheduling and intra-node scheduling for divisible loads is presented on the heterogeneous cluster systems of multi-core computers, which have different number of processing cores and distinct computation, communication and memory abilities. The presented inter-node scheduling can accommodate dynamicaly the surplus of the remaining available memory of nodes in each round scheduling, and the intra-node scheduling can efficiently make use of the characterictics of multi-core machines and the shared L2 cache to reduce the required time to access memory. The experiment results with different tests and several parameters on a heterogeneous cluster system with multi-core computers show that the presented algorithm can obtain better scheduling performance and reduce significantly the total scheduling length, and it can schedule more large-scale applications than the existing scheduling scheme.