Skip to Main content Skip to Navigation
Theses

Distribution dynamique adaptative à l'aide de mécanismes d'intelligence collective

Abstract : This work presents a dynamic adaptive distribution method for distributed applications made of large number of interacting entities in a versatile computation environment. Load balancing as well as communication minimization are taken into account. The proposed method is based on the detection of organizations inside the application to better distribute it. Organizations are identified as groups of highly communicating entities. Organizations evolve, appear, strengthen, weaken and disappear. Available computing resources where the application runs also change. Such constraints dictate that the distribution be dynamic and adaptive. The method is based on colonies of numerical ants trying to gather entities of the application. Ants cooperate inside a unique colony and compete when they are not in the same colony. They try to capture organizations inside the application, each colony working for a distinct computing resource. Competition between colonies allow the load balancing. Collaboration inside colonies allow to detect organizations, putting highly communicating sets on the same computing resource. Finally, population management allow to take into account computing resources heterogeneity.
Complete list of metadatas

Cited literature [75 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00453910
Contributor : Antoine Dutot <>
Submitted on : Friday, February 5, 2010 - 9:05:30 PM
Last modification on : Tuesday, February 5, 2019 - 11:44:22 AM
Long-term archiving on: : Thursday, October 18, 2012 - 2:40:38 PM

Identifiers

  • HAL Id : tel-00453910, version 1

Citation

Antoine Dutot. Distribution dynamique adaptative à l'aide de mécanismes d'intelligence collective. Modélisation et simulation. Université du Havre, 2005. Français. ⟨tel-00453910⟩

Share

Metrics

Record views

534

Files downloads

772