Evaluation of Deployment Models for Microservices in an Edge-based Serverless Environment

In the edge computing ecosystem, microservices and serverless computing are gaining attention recently. The first one provides a highly flexible and scalable software design, while the latter offers a resource efficiency mechanism for running software, promising for IoT deployment at the edge. However, microservices, mainly deployed by Kubernetes (K8s) – a container orchestration platform, has intrinsic overhead, which is potentially exacerbated by the container/pod layer introduced by K8s. This problem, while can be negligible in the cloud, might be significantly limited in terms of capabilities in edge environments. To this end, this article investigates the impacts of different deployment types, namely native, container, and pod, for microservices in various categories: resource consumption, performance, availability, and scalability. The results show quantitative measurements of the mentioned metrics and, in particular, also the advantages and disadvantages of each deployment type for developers to design their microservices efficiently.