An Efficient Microservices Architecture for MLOps

Abstract

In a microservices architecture, each service has a database. Hence, it is important to communicate and synchronize data between services. The SAGA pattern is a traditional microservices architecture pattern, and the command query responsibility segregation (CQRS) pattern has recently attracted increasing attention. Machine learning model operation management (MLOps) aims to stably deploy and maintain the system by preprocessing big data and learning machine learning models. Data processing in the microservices architecture is important because considerable data is used. This paper proposes an appropriate architecture for each microservice to perform efficiently in the MLOps environment.