Improvement of container scheduling for Docker using Ant Colony Optimization

Docker, a software container implementation, has emerged not only as an operating-system level virtualization but also an application delivery platform for today Internet. However, the scheduling algorithm shipped with SwarmKit, the orchestration engine behind Docker, is suboptimal when resources are nonuniform. The use of meta-heuristic, like Ant Colony Optimization (ACO), is feasible to improve the scheduler's optimality. This paper presents a study of ACO to implement a new scheduler for Docker. The main contribution of this paper is an ACO-based algorithm, which distributes application containers over Docker hosts. It is to balance the resource usages and finally l eads to the better performance of applications. The experimental results showed that workloads placed by ACO performed better than those of the greedy algorithm by approximately 15% on the same host configuration.