Dynamic load-balancing via a genetic algorithm

We produce a genetic algorithm (GA) scheduling routine, which with often relatively low cost finds well-balanced schedules. Incoming tasks (of varying durations) accumulate, then are periodically scheduled, in small batches, to the available processors. Two important priorities for our scheduling work are that loads on the processors are well balanced, and that scheduling per se remains cheap in comparison to the actual productive work of the processors. We also include experimental results, exploring a variety of distributions of task durations, which show that our scheduler consistently produces well-balanced schedules, and quite often does so at relatively low cost