Managing Microservices For Open Systems
Zhongyi Lu, Declan T. Delaney and David Lillis
SCIENCE CHINA Information Sciences, 2026.
Abstract
The evolution of the digital era has increased the need for open microservice systems that are intelligent and allow arbitrary microservice behaviour. This requires systems to have a strong ability to manage microservices. A common microservice management strategy is allocating microservices into domains. Traditional domain-management strategies, which group microservices by predefined criteria, fail in these open environments as they require prior system knowledge. To address this, we propose a novel domain-management framework that autonomously constructs "Similarity Domains" based on microservice functionality. The framework computes the functional similarity between microservices and builds domains that store similar microservices together. Our framework first employs Natural Language Processing (NLP) techniques to generate semantic embeddings from OpenAPI documentation, effectively capturing functional semantics. These embeddings are then clustered to form domains of functionally similar microservices without any prior knowledge, enabling resilient service discovery and seamless integration of newcomers. Novel evaluation methods were designed for the proposed framework. Evaluations conducted on real-world OpenAPI documents, as well as through simulations of an open microservice system, demonstrate that the proposed approach performs effectively compared to existing domain-management strategies, facilitating appropriate service discovery, domain formation, and integration of newcomer microservices.