Dependency-aware Microservice Deployment and Resource Allocation in Distributed Edge Networks

Abstract

Microservice deployment strategy gains increasing attentions due to its benefits for reducing the overall computation and transmission delay of applications. Most prior work models the dependencies between microservices of an application as service function chain or directed acyclic graph, and neglects the interaction relationship and the different communication overhead between two interactive microservices. To characterize the features of applications, our work considers the interaction dependency between microservices, and models the communication frequency over each dependency edge. In this paper, we address the joint optimization of microservice deployment and resource allocation decisions to minimize the sum of delay of multiple applications in the distributed edge networks. Since this problem is NP-hard, a coalition game-based algorithm is proposed to obtain the dependency-aware microservice deployment and resource allocation scheme. Numerical results validate the advantages of our approach particularly when microservices have strong interaction.