Optimization of Microservice Composition Based on Artificial Immune Algorithm Considering Fuzziness and User Preference

Microservices is a new paradigm in cloud computing that separates traditional monolithic applications into groups of services. These individual services may correlate or cross multi-clouds. Compared to a monolithic architecture, microservices are faster to develop, easier to deploy, and maintain by leveraging modern containers or other lightweight virtualization. To satisfy the requirements of end-users and preferences, appropriate microservices must be selected to compose complicated workflows or processes from within a large space of candidate services. The microservice composition should consider several factors, such as user preference, correlation effects, and fuzziness. Due to this problem being NP-hard, an efficient metaheuristic algorithm to solve large-scale microservice compositions is essential. We describe a microservice composition problem for multi-cloud environments that considers service grouping relations and corresponding correlation effects of the service providers within intra- or inter-clouds. We use the triangular fuzzy number to describe the uncertainty of QoS attributes, the improved fuzzy analytic hierarchy process to calculate multi-attribute QoS, construct fuzzy weights related to user preferences, and transform the multi-optimal problem into a single-optimal problem. We propose a new artificial immune algorithm based on the immune memory clone and clone selection algorithms. We also introduce several optimal strategies and conduct numerical experiments to verify effects and efficiencies. Our proposed method combines the advantages of monoclone, multi-clone, and co-evolution, which are suitable for the large-scale problems addressed in this paper.