On Exploiting Edge Resources for Micro-Service Based SaaSs

Abstract

Many popular Software as a Services (SaaSs) have stringent latency requirements that have necessitated the use of edge resources. Many SaaSs use micro-services-based architecture where the SaaSs are designed as a set of independent smaller services known as micro-services. In this work, we model a SaaS comprising multiple micro-services using a directed graph, wherein each node represents a micro-service and each directed edge represents a possible navigation path for a user from one micro-service to another. We consider a network with multiple servers at varying distances from the end-users that can be used to host these micro-services. The evolution of this system consisting of servers, SaaS, and its users is then mapped to an open feed-forward queueing network. We formulate an optimization problem to identify the server to host each micro-service and to determine the amount of computational resources to rent at these servers. Structural properties of the optimal solution to this problem are characterized under some assumptions on the graph structure and relative values of the costs involved. In addition, we propose a heuristic for solving the formulated problem and illustrate its efficacy via numerical results.