Task and resource assignment in distributed real-time systems

Abstract

In this paper, we propose an efficient algorithm to statically assign tasks and logical resources to the processors, taking into account the blocking times due to resource contention. We formulate the assignment problem using an assignment graph where the weights on the edges of the graph represent the desirability of assigning the tasks (connected by the edges) to the processors. These weights are computed based on the resource usages, communication costs, and the priorities of the tasks. An efficient clustering and graph-bisection algorithm is used to find an assignment of the tasks and resources such that the total communication costs and total blocking time are minimized, subject to the constraint that the tasks are schedulable after the assignment. Experimental results indicate the effectiveness of our algorithm.<>