Multidimensional Resource and Load Collaborative Scheduling Algorithm Based on Reinforcement Learning for Cloud Data Centers

Task scheduling for multi-dimensional resources is one of the most fundamental problems in cloud data centers (DC). Among existing resolutions, the Q-learning model has been considered an excellent tool for fast task scheduling in DC environments. In this paper, we propose a load-balancing model for multi-dimensional resource scheduling in a cloud DC and a Q-learning based task scheduling algorithm (TSQL) that aims to reduce task makespan time and improve resource utilization. Simulation results show that, compared with existing algorithms, our algorithm optimizes 46.92%, 33.67% in makespan and resource utilization.