Adaptive Auto-Scaling in Serverless Computing

Dr. Sudesh Rani

Abstract: Serverless computing has transformed modern cloud application deployment by removing infrastructure management responsibilities from developers. However, dynamic workloads create challenges such as cold start latency, resource allocation inefficiency, and operational cost optimization. Adaptive auto-scaling addresses these challenges by dynamically adjusting resources according to workload demand. This paper reviews adaptive auto-scaling techniques in serverless computing, including reactive, predictive, reinforcement learning-based, and hybrid scaling approaches. It also proposes an Intelligent Hybrid Adaptive Scaling Framework (IHASF) that integrates workload prediction, reinforcement learning, and warm container management to improve Quality of Service (QoS), reduce latency, and optimize resource utilization. The study further discusses existing challenges, performance metrics, and future research directions for adaptive scaling in serverless environments.

Keywords: Serverless Computing, Auto-Scaling, Cloud Computing, Function-as-a-Service, Reinforcement Learning, Resource Allocation