Skip to content
Conference

Reinforcement Learning Based SLA-Aware Resource Allocation for Cost-Efficient Cloud Computing

Sep 2026 · 2026 International Conference on Secure IoT and Cybersecurity (ICOSICS) · pp. 1543-1548 · 0 citations · 22 references

Abstract

Inefficient resource allocation has been posing a significant problem, leading to increased operational costs and violations of SLAs. Effective scheduling of users' jobs while balancing availability, response time delay and throughput is difficult. This paper proposes an automated resource allocation framework using DQN based Reinforcement Learning (RL) agent and optimal allocation of resources. The system proposed framework runs inside a Kubernetes environment that will monitor cloud workload conditions, such as a virtual machine's CPU usage and response time, in order to make decisions according to the system'. By using the DQN method, the experimented architecture has demonstrated that it has increased resource utilization by 23% and SLO violations by 31%. Typically, traditional allocation schemes are dependent on either static provisioning or auto-scaling through rules. Even though these methods are commonly found in current cloud systems, they frequently fail to adapt well to ever-changing and intricate workload patterns. Static provisioning can result in resource depletion during low demand periods and a shortage of resources during high workloads. Thus, the challenge of achieving SLA compliance, Quality of Service and decreasing operational expenses in modern cloud infrastructures has become a major research problem.

View source

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.