Multi-dimensional Resource Allocation in Distributed Data Centers using Deep Reinforcement Learning

With the development of edge-cloud computing technologies, distributed data centers (DCs) have been extensively deployed across global Internet. Since different users/applications heterogeneous requirements on specific types ICT resources in DCs, how to optimize such under dynamic and even uncertain environments becomes a challenging issue. Traditional approaches are not able provide effective solutions for multi-dimensional resource allocation that involves balanced utilization DC environments. This paper presents reinforcement learning based approach (termed as NESRL-MRM) is achieve availability To train NESRL-MRM’s agent with sufficiently quick wall-clock time but without loss exploration diversity search space, natural evolution strategy (NES) employed approximate gradient reward function. realistically evaluate performance NESRL-MRM, our simulation evaluations real-world workload traces from Amazon EC2 Google datacenters. Our results show NESRL-MRM significant improvement over existing balancing resources, which leads substantially reduced blocking probability future incoming demands.