Abstract
This paper proposes a reinforcement learning-based method for microserviceresource scheduling and optimization, aiming to address issues such as unevenresource allocation, high latency, and insufficient throughput in traditionalmicroservice architectures. In microservice systems, as the number of servicesand the load increase, efficiently scheduling and allocating resources such ascomputing power, memory, and storage becomes a critical research challenge. Toaddress this, the paper employs an intelligent scheduling algorithm based onreinforcement learning. Through the interaction between the agent and theenvironment, the resource allocation strategy is continuously optimized. In theexperiments, the paper considers different resource conditions and loadscenarios, evaluating the proposed method across multiple dimensions, includingresponse time, throughput, resource utilization, and cost efficiency. Theexperimental results show that the reinforcement learning-based schedulingmethod significantly improves system response speed and throughput under lowload and high concurrency conditions, while also optimizing resourceutilization and reducing energy consumption. Under multi-dimensional resourceconditions, the proposed method can consider multiple objectives and achieveoptimized resource scheduling. Compared to traditional static resourceallocation methods, the reinforcement learning model demonstrates strongeradaptability and optimization capability. It can adjust resource allocationstrategies in real time, thereby maintaining good system performance indynamically changing load and resource environments.