|
 GRAPPA
GRid-based Advanced Programming Paradigms and Architectures
In this research theme, we investigate how GRID or other distributed computing organizations can be efficiently managed by introducing self-organization and self-adaptation. The theme breaks down in the following concrete research projects:
- ADAPT: In this project, we study the different adaptation mechanisms that can be used to introduce self-modification. We distinguish between self-adaptation, external adaptation and outsourcing. Starting point for studying these mechanisms is the definition of a Minimal Agent. By this we mean the most elementary structure that constitutes the minimum for an agent to exist. Whenever required, this skeleton can then be extended by adding functional components (e.g. navigation, negotiation,...). How and when to incorporate such components are the focus of the research.
- SCALA: In this project, we study ways of improving the management of distributed computing systems. The main idea is to introduce self-awareness and self-adaptation mechanisms. More specifically, we propose a simple though efficient mechanism to enable distributed processing among nodes to communicate with each other and to dynamically rebalance the workload among them. This may be needed for several reasons. Local congestion or a node failure may induce the necessity to re-allocate tasks to other nodes. Task reallocation means that the snapshots of the current work load is available and can be used to perform efficient re-allocation. The question arises as to how this allocation should be done by the co-allocating agents. Different approaches and solutions have been proposed to tackle this intriguing problem.
If you are interested in participating in this research, do not hesitate to contact me.
|
