DEEP REINFORCEMENT LEARNING FOR SCALABLE TASK SCHEDULING IN SERVERLESS COMPUTING
Unknown authors
Abstract
Serverless computing has revolutionized the way applications are deployed and managed by abstracting server management and enabling scalable resource allocation.However, efficient task scheduling remains a critical challenge to optimize performance and resource utilization.This paper explores the application of Deep Reinforcement Learning (DRL) to develop a scalable task scheduling framework tailored for serverless environments.We leverage the publicly available AWS Lambda dataset, which provides real-world workload traces, to train and evaluate our DRL model.Our approach dynamically adjusts scheduling decisions based on real-time workload patterns, leading to improved response times and reduced resource wastage compared to traditional scheduling algorithms.Experimental results demonstrate that the DRL-based scheduler outperforms baseline methods by achieving a 15% increase in task throughput and a 20% decrease in average latency.This study highlights the potential of DRL in enhancing the efficiency and scalability of serverless computing platforms, paving the way for more intelligent and adaptive resource management strategies.