A GENERIC SMA FOR MULTI CRITERIA ALLOCATION OF APPLICATIONS ON HETEROGENEOUS ARCHITECTURES

We are interested in this work, by the optimized static allocation multi criteria in a real time

distributed system when the tasks are subject to precedence’s constraints. Within this

framework it is necessary to suppose before execution, that all the possible scenarios of

executions satisfy the temporal constraints while minimizing the cost and the size of material

architecture, as well as the use at best of its resources. In this problem of resource allocation

(placement and scheduling), which is NP-Complete, the satisfaction of several criteria can be

contradictory. For the resolution of this problem, we propose in this work, a generic multi

agent system, which is a dynamic component, coupled with a static strategy of scheduling in

order to particularly integrating the criterion of load balancing. Thus, the need for a dynamic

model appeared to us with the consideration of the heuristic based on list scheduling [1], [12].

An experimental Analysis was realized under programming parallel environment PVM

(Parallel Virtual Machine)}, and shows the interest of our method. This for any heuristics

using the dates of execution of the tasks (operations) in particular for method AAA

(Algorithm, Architecture, Adequacy) developed with the INRIA and which was the subject of

several extensions.