Latency aware graph-based microservice placement in the edge-cloud continuum

The deployment of cloud-based services on multi-tier edge cloud infrastructure is a research problem for application providers. A microservice-based architecture plays a vital role in addressing cloud-based service placement by dividing the monolithic application into microservices. Latency-aware placement is another important means of reducing the overall response time of microservice-based applications. Considering the placement problem, this paper's contribution is threefold; firstly, a graph partition technique, the Leiden algorithm, for community detection, and an Integer Linear Programming model is proposed for microservices-based applications. Secondly, we orchestrate various microservice-based applications through lightweight Kubernetes distribution with distinct, realistic federated edge cloud architecture to prove the importance of how the internal interaction among microservices can affect the overall performance. Lastly, the results section shows the difference between the community detection algorithms and reports latency and computation time at the edge-cloud continuum.