Utilizing Deep Reinforcement Learning for Resource Scheduling in Virtualized Clouds

Cloud computing can transform a large amount of computer technology into services where computer resources are virtualized and made available as utility services. Therefore, it is essential to schedule virtual resources for maximum energy consumption and cost utilization. According to the sustainable development goals announced by United Nations, SDG7 calls for enhancing energy consumption and power efficiency. Thus, Machine learning (ML) can be used to schedule resources efficiently or optimal request allocation. Deep reinforcement learning (DRL), a combination of deep learning (DL) and reinforcement learning (RL), is one area of ML and has a considerable prospect in resource scheduling of cloud computing. This paper introduces a DRL-based scheduling algorithm in cloud computing. The algorithm aims at optimizing dynamic resource allocation for large-scale cloud computing environments. With the architecture of DLR, the proposed model adopts the mapping process to analyze the environment and construct the final scheduler gradually. The algorithm considers reducing the task response and maintaining a satisfying Quality of Service while reducing energy consumption as envisaged in SDG 7.