The objective of this thesis will be to find the solutions for Scheduling and Deployment adapted to an environment of grid computing.
Scheduling in heterogeneous medium
The very first problem with which we are confronted relates to applications scheduling. It is significant to be able to find the computing power and the capacity storage necessary to a task; it remains with a given task. Once identified the server able to solve the task, it remains to determine a scheduling that offers the greatest possible effectiveness. The installation of such mechanisms in a heterogeneous medium of this scale initially requires lying out the information of two types:
the necessary resources can be obtained either by an approach based on algorithm modeling (if possible) or by a statistical approach based on a series of executions.
the available resources, for example with the installation of sensors giving a state of the system.
The study and the setting open of these two approaches were initiated within project DIET (Distributed Interactive Toolbox Engineering), but of much work remain to develop before validating these results.
Deployment and scheduling in hierarchical medium
The concept of network hierarchical structure the Grid, implies strong constraints relating to extensibility of the platforms of grid computing. A solution consists in considering a hierarchical approach of the executive support offering to the concept client-agent-server, an advanced structure multi-agents. However, that introduces the significant problems of deployment. In effect, the placement of the various agents becomes crucial in term of performance.
It is then particularly interesting to be able to offer tools for deployment in adequacy with the scheduling can put at contribution.