Grid'5000 experiment

Jump to: navigation, search

Towards self-healing Web service-base applications. (Application)

Conducted by

Riadh Ben-halima, Ismael Bouassida-rodriguez


Internet progress has enabled data exchange between remote collaborators and the multiplication of on-line services. Several platforms of services are provided today to support the design, the deployment and the implementation of these services in reduced scales of time. However, users require more exigencies. They dislike to encounter problems while using a service. For example, wait for a long time in order to book a room, check whether the crashed rent car service is restored, search for services with improved QoS. Therefore, we have to deal with these problems at the design time, and provide a strategy for recovery in order to satisfy users exigencies. We address QoS monitoring and measurement of a Web service-based application. We are achieving experiments on the cooperative reviewing system. We interest mainly in the conference search scenario. The measurement is carrying out while using a connector based approach. We present the performance measurement values and analyze experiment results. The measurement is fulfilled under the grid5000 which is an experimental grid platform gathering more then 3000 nodes over eight geographically distributed sites in France.


in progress


  • Nodes involved: 500
  • Sites involved: >3
  • Minimum walltime: 1h
  • Batch mode: yes
  • Use kadeploy: no
  • CPU bound: no
  • Memory bound: no
  • Storage bound: no
  • Network bound: no
  • Interlink bound: no

Tools used

JDK1.5 + TOMCAT5.9 + MYSQL5 + AXIS1.3


In this stage, we have dressed curve for the monitored QoS parameters. The first exploitation concerns the variation of QoS while varying client number. This will be used to diagnose current running web service, in order to recover QoS degradation. The experiments allows us to measure how QoS vary under many load conditions, and how to prevent service from QoS degradation. For Web service based application, the monitors load come to 0,5 second when we exceed 50 requesters. Also, when we exceed 50 requesters, the service availability turns down to less than 50%. Consequently, we have to limit the number of concurrent requests addressed to a service to less than 50 requesters. The experiment shows the feasability and the efficiency of the self-healing architecture that we developed in the context my Phd thesis.

Shared by: Riadh Ben-halima, Ismael Bouassida-rodriguez
Last update: 2009-10-28 15:43:10
Experiment #445