iScheduler: Reinforcement Learning-Driven Continual Optimization for Large-Scale Resource Investment Problems

Researchers have developed iScheduler, a reinforcement learning system that optimizes large-scale task and resource scheduling for cloud platforms. It formulates the complex scheduling problem as a sequence of decisions, allowing it to find workable solutions much faster than traditional programming methods. In tests on a new industrial benchmark with up to 10,000 tasks, it achieved competitive resource costs while reducing the time to find a feasible schedule by up to 43 times.