Dynamic Resource Allocation via Deep Q-Networks for Efficient Load Balancing and Task Scheduling in Cloud Environments

Abstract

Cloud computing is the domain that enables task scheduling to carry out complex jobs more efficiently is load balancing. In this context, one of the most important tasks in task scheduling and resource assignment is load balancing. Existing load balancing methods have a hard time predicting the sharp decline in workloads that leads to overloading or increased resource usage. These challenges are addressed by proposing the improved load balancing, namely Gradient-Based Load balancing and Cognitive Deadline Shaping Scheduler (GBLB-CDSS). The load balance of the nodes in the GBLB-CDSS is radially. The burning mode is to schedule task deadlines with correct routing decisions. In this paper, the IDN is proposed to solve GBlb-CDSS. One such learning scheme is the IDN, which is a RL-based solution that finds good policies for traffic patterns with an RL agent taking decisions on-line period by period. The experimental evaluation is carried out using the CloudSim Simulation Dataset. Simulations indicate that the proposed scheme achieves a superior performance in Average Response Time (ART), Missed Deadline Ratio (MSR), Task Rejection Rate (TRR), Average VM Usage (AVM), and Load Variance (LV).