Scheduling Microservice-based Workflows to Containers in On-demand Cloud Resources

摘要

Though microservices process and communicate with lightweight mechanisms, finer tasks result in much more complicated precedence constraints. Different tasks have distinct resource requirements and different VMs (Virtual Machine) have various configurations and prices. In this paper, we consider the problem of scheduling microservice tasks of workflow applications to containers configured on on-demand VMs to minimize the total rental cost. The problem is mathematically modelled using integer programming and an algorithm framework is proposed. For dynamic available containers and resource requirements, a task scheduling heuristic is presented for scheduling precedence-constrained or independent tasks to available containers. All parameters and components of the proposed algorithm framework are statistically calibrated by the Analysis of Variance technique on a large number of random instances. Performance of the the proposed algorithm is verified over a lot of instances.