Performance analysis of multiple site resource provisioning: effects of the precision of availability information

  • Authors:

  • Marcos Dias De Assunção;Rajkumar Buyya

  • Affiliations:

  • Grid Computing and Distributed Systems Laboratory, Department of Computer Science and Software Engineering, The University of Melbourne, Australia and NICTA Victoria Research Laboratory, The Unive ...;Grid Computing and Distributed Systems Laboratory, Department of Computer Science and Software Engineering, The University of Melbourne, Australia

  • Venue:
  • HiPC'08 Proceedings of the 15th international conference on High performance computing
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Emerging deadline-driven Grid applications require a numberof computing resources to be available over a time frame, startingat a specific time in the future. To enable these applications, it is importantto predict the resource availability and utilise this informationduring provisioning because it affects their performance. It is impracticalto request the availability information upon the scheduling of everyjob due to communication overhead. However, existing work has notconsidered how the precision of availability information influences theprovisioning. As a result, limitations exist in developing advanced resourceprovisioning and scheduling mechanisms. This work investigateshow the precision of availability information affects resource provisioningin multiple site environments. Performance evaluation is conductedconsidering both multiple scheduling policies in resource providers andmultiple provisioning policies in brokers, while varying the precision ofavailability information. Experimental results show that it is possible toavoid requesting availability information for every Grid job scheduledthus reducing the communication overhead. They also demonstrate thatmultiple resource partition policies improve the slowdown of Grid jobs.