A Heuristic Approach to Task Allocation in Real- Time Distributed Systems

The task scheduling problem consists in defining a time sharing of system resources that can meet all logical and temporal constraints of the tasks being considered. In real-time distributed systems such a problem can be broken up in two related steps: task allocation and local task scheduling. Task allocation can be performed dynamically or statically. For tasks with predictable behavior (periodic or sporadic), static scheduling provides a reliable way to meet time constraints. This is of fundamental importance when defining an allocation for hard-deadline tasks. Unfortunately, finding an optimal feasible allocation is known to be NP-hard. Therefore, heuristic approaches for finding good allocations in polynomial time have been studied. In this paper we describe the use of a new heuristic to generate static schedules for critical periodic tasks in real-time distributed systems. In order to evaluate the benefits of the proposed approach, termed microcanonical optimization, a study is performed to compare the quality of the achieved schedules with the ones obtained with alternative heuristics, namely, simulated annealing, tabu search and Tsallis’s annealing. Our experiments show that microcanonical optimization is able to outperform the competing heuristics, achieving better allocations in shorter processing times.