Multi-Objective Multi-Factorial Evolutionary Algorithm for Container Placement

The use of containerization technology in microservice architecture has become widespread, owing to its potential to support fast deployment of web applications and improve the resource utilization in cloud data centers. Evolutionary algorithms (EAs) have been performed promising on the deployment of applications created using microservices. However, with the growing demand for microservice application, the existing EAs fail to solve the large-scale container placement problem due to the high time complexity and poor scalability. A multi-factorial evolutionary algorithm (MFEA) is proposed in this article, which can evolve multiple optimization problems simultaneously for the container placement problem in heterogeneous cluster environments. First, a system model integrated the heterogeneous clusters, microservices, containers, and four optimization objectives is presented. Then, embedded with local search strategy, a multi-objective container placement MFEA (MOCP-MFEA) algorithm is developed to address the container placement problem. MOCP-MFEA is applied to a variety of container placement problems with different application sizes in heterogeneous cluster environments. Experimental results show that compared with various conventional and evolutionary-based approaches, MOCP-MFEA could shorten optimization time significantly and offer a competitive placement solution for the container placement problem and show a good elasticity in heterogeneous cluster environments. Moreover, the deployment scheme of container to physical machines is crucial to lowering resources wastage.