Academic Journals Database
Disseminating quality controlled scientific knowledge

Agentless robust load sharing strategy for utilising hetero-geneous resources over wide area network

Author(s): Natthakrit Sanguandikul

Journal: Maejo International Journal of Science and Technology
ISSN 1905-7873

Volume: 5;
Issue: 02;
Start page: 215;
Date: 2011;
VIEW PDF   PDF DOWNLOAD PDF   Download PDF Original page

Keywords: load sharing strategies | self-scheduling strategies | high performance computing | distributed systems | heterogeneous systems | load balancing | task assignment

Resource monitoring and performance prediction services have always been regarded as important keys to improving the performance of load sharing strategy. However, the traditional methodologies usually require specific performance information, which can only be collected by installing proprietary agents on all participating resources. This requirement of implementing a single unified monitoring service may not be feasible because of the differences in the underlying systems and organisation policies. To address this problem, we define a new load sharing strategy which bases the load decision on a simple performance estimation that can be measured easily at the coordinator node. Our proposed strategy relies on a stage-based dynamic task allocation to handle the imprecision of our performance estimation and to correct load distribution on-the-fly. The simulation results showed that the performance of our strategy is comparable or better than traditional strategies, especially when the performance information from the monitoring service is not accurate.
Why do you need a reservation system?      Save time & money - Smart Internet Solutions