Dynamic Distributed Scheduling of Soft Real-Time Tasks

In distributed real-time systems the scheduler of an overloaded node may choose to transfer the execution of one or more of its tasks to other less busy nodes in order to improve the number of tasks that meet their deadlines. One solution uses Bayesian theory to infer the loading state of the system and, based on this information, the scheduler chooses an “adequate” node to transfer a task to. The concept of “adequate” will be a function of objectives that are established in the location policy adopted. In the Bayesian decision method, objectives are represented by an utility function (U): given a system state, U returns, for each possible decision to be taken, a number that represents the level of accordance with the required objectives. However, the development of a utility function can be a tricky and somewhat subjective task. In this paper, we describe a new method which easily maps transfer objectives into useful mathematical expressions, representing location policy objectives by fuzzy sets. The proposed method was employed to add objectives to a Bayesian decisionbased algorithm aiming to improve the number of aperiodic tasks executed over time in a distributed real-time system.