Load profiling: a methodology for scheduling real-time tasks in a distributed system

Traditionally, the goal of load management protocols for distributed systems has been to ensure that nodes are equally loaded. We show that for real time systems, load balancing is not desirable since it results in the available bandwidth being distributed equally amongst all nodes-in effect making all nodes in the system capable of offering almost the same bandwidth (e.g., in cycles per second) to incoming tasks. We show that this “one size fits all” practice leads to a higher rate of missed deadlines as incoming tasks may be denied service because they require bandwidth that cannot be granted at any single node-while plenty of fragmented bandwidth is collectively available in the system. We propose a new load profiling strategy that allows the nodes of a distributed system to be unequally loaded so as to maximize the chances of finding a node that would satisfy the computational needs of incoming real time tasks. The performance of the proposed protocol is evaluated via simulation, and is contrasted to other dynamic scheduling protocols for real time distributed systems