Skip to Main content Skip to Navigation

Introduction de fonctionnalités d'auto-optimisation dans une architecture de selfbenchmarking

Abstract : Benchmarking client-server systems involves complex, distributed technical infrastructures, whose management deserves an autonomic approach. It also relies on observation, analysis and feedback steps that closely matches the autonomic control loop principle. While previous works in performance testing have shown how to introduce autonomic load testing features through self-regulated load injection, the goal of this thesis is to follow this approach of autonomic computing to introduce self-optimization features in this architecture to obtain reliable and comparable benchmark results, and to achieve the fully principle of Self-benchmarking.Our contribution is twofold. From the algorithmic point of view, we propose an original optimization algorithm in the context of performance testing. This algorithm is divided into two parts. The first one concerns the overall level, i.e. the control of the performance index evolution, based on global parameters setting of the system. The second part concerns the search for the optimum when only one parameter is modified. From the software architecture point of view, we complete the Fractal component-based architecture, containing several autonomic control loops (saturation, injection, optimization computing) and we implement the coordination principle between these loops by asynchronous messages according to the publish-subscribe communication paradigm. To apply a given parameters setting on the system under test, we introduced new components Configurators to support the setting of parameters before starting the test process. It may also be necessary to restart all or part of the system to optimize to ensure that the new setting is effectively taken into account. We introduced components Starters to cover this need in a specific way for each system.To validate our self-optimization framework, two types of campaigns have been conducted onto the servers of Orange Labs in Meylan and the servers of the LISTIC Laboratory of the University of Savoie in Polytech Annecy-Chambéry (Annecy le Vieux). The first one is a WEB online shopping application deployed on a Java EE application server JonAS. The second one is a three-tiers application (WEB, business (EJB JOnAS) and data base) clusterSample. The three tiers are in three separate machines.
Document type :
Complete list of metadata

Cited literature [67 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Tuesday, January 29, 2013 - 1:17:10 PM
Last modification on : Friday, November 6, 2020 - 3:36:59 AM
Long-term archiving on: : Tuesday, April 30, 2013 - 4:15:18 AM


Version validated by the jury (STAR)


  • HAL Id : tel-00782233, version 1



El Hachemi Bendahmane. Introduction de fonctionnalités d'auto-optimisation dans une architecture de selfbenchmarking. Autre [cs.OH]. Université de Grenoble, 2012. Français. ⟨NNT : 2012GRENA020⟩. ⟨tel-00782233⟩



Record views


Files downloads