Microservice Aging and Rejuvenation

Due to ‘aging’, not only the service rate of the software decreases with time but the software itself experiences occasional crash/hang failures. Software rejuvenation involves occasional stopping the executing software, ‘cleaning’ the ‘internal state’ and restarting. Moreover, container technology promotes the process of fair and efficient allocation of physical resources among virtual machines. However, the emergence of distributed cloud platform undoubtedly increase the flexibility and complexity of the system. In this paper, we propose a method of predicting microservice [1] aging by deep learning, and a rejuvenation policy by the CVA architecture. From the perspective of container, it can provide a technology of vertical expansion and contraction of container resources, improve the utilization of resources in the clustered environment, and improve the availability of microservice system.