Dynamic Decomposition of Service Function Chain Using a Deep Reinforcement Learning Approach

The Internet of Things (IoT) universe will continue to expand with the advent sixth generation mobile networks (6G), which is expected support applications and services higher data rates, ultra-reliability, lower latency compared fifth (5G). These new demanding 6G introduce heavy load strict performance requirements on network. Network Function Virtualization (NFV) a promising approach handling these challenging requirements, but it also poses significant Resource Allocation (RA) challenges. Especially since network be highly complicated comparatively short-lived, operators compelled deploy in flexible, on-demand, agile manner. To address aforementioned issues, microservice approaches are being investigated, decomposed loosely coupled, resulting increased deployment flexibility modularity. This study investigates RA for microservices-based NFV efficient decomposition Virtual (VNF) onto substrate networks. VNFs involves additional overheads, have detrimental impact resources; hence, finding right balance when how much allow critical. Thus, we develop criterion determining potential/candidate granularity such decomposition. joint problem embedding microservices model solve using exact mathematical models. Therefore, implemented Reinforcement Learning (RL) Double Deep Q-Learning, revealed an almost 50% more normalized Service Acceptance Rate (SAR) over monolithic VNFs.