Adaptive Cache-Driven Request Distribution in Clustered EJB Systems

  • Authors:
  • Hazem Elmeleegy;Noha Adly;Magdy Nagi

  • Affiliations:
  • Mubarak City for Scientific Research, Egypt;Alexandria University, Egypt;Alexandria University, Egypt

  • Venue:
  • ICPADS '04 Proceedings of the Parallel and Distributed Systems, Tenth International Conference
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper presents a new algorithm for requestdistribution in clustered EJB systems. A new classificationfor EJB requests is first introduced, on which requestdistribution is based. The objective is to achieve loadbalancing and to enhance the caching performance in theEJB containers. The algorithm is based on periodicallycollecting traffic statistics from the EJB containers, andthen solving a constrained optimization problem that findsthe best distribution strategies for each request type. Anevent-driven simulator is built for the performanceevaluation, where the TPC-W benchmark is used as theworkload. Simulation results show that for relatively smallsizes of the containers' caches, the new algorithmoutperforms the random and round robin algorithms,currently used in clustered EJB systems. In addition, itscales better with the cluster size and can adapt to thevarying load patterns.