Bayesian Optimization-Based Task Scheduling Algorithm on Heterogeneous System

In heterogeneous computing systems, efficient task scheduling is essential for utilizing resources and reducing computing time. This problem has been shown NP-complete in the general case. Existing solutions are mainly heuristic-based that would easily track into optimal local solutions and reinforcement learning-based that need an expensive computation cost for data training on neural networks. To overcome the shortcomings, we propose a Bayesian optimization based task scheduling algorithm that automatically searches for the best heuristic strategy in the problem space. Our algorithm builds a Bayesian optimization model on heuristic strategy and scheduling performance, and updates the model by interacting with the environment to find the optimal solutions globally. To enhance the confidence of our experiments, we measure the average (weighted) makespans and running time of our algorithm. The experimental results show that our approach can improve the scheduling performance compared to the baselines.