Dynamic Microservice Deployment and Offloading for Things-Edge-Cloud Computing

The growing edge cloud computing paradigm allows flexible handling of latency-sensitive and computation-intensive applications operating on user devices as the Internet of Things and 5G technologies gain in popularity. Microservices based on container technology are regarded as a potential architecture when applied to edge computing because of their lightweight and layered image properties. However, many current studies on the combination of the two simply treat microservices as a replacement for traditional virtual machine architecture without fully utilizing its advantages. In addition to discussing the impact of image loading strategy on neighboring time slots, this paper also focuses on the advantages of microservices layered image sharing. Our research in this paper studies the microservice deployment and task offloading of a mobility-aware things-edge-cloud system, and a deep reinforcement learning-based algorithm is proposed in this work to make decisions that optimize the system’s long-term throughput and delay utility.