Context-Aware and QoS Prediction-based Cross-Domain Microservice Instance Discovery

Abstract

Context and QoS cannot be fully used and it is difficult to balance the time efficiency and the accuracy in most existing microservice instance discovery algorithms. So, a method for microservice instance discovery based on context clustering and QoS prediction is proposed. Firstly, according to the similarity of microservice context and user context, clustering algorithms are applied to match requesters with microservices. Then the current attribute data is predicted according to the history data, which includes instance QoS data and server resource data that instances are located on. Finally, the performance value of each instance can be got and instances with the highest performance value are returned to the requester. The comparison experiments show that the microservice instance discovery method based on context clustering and QoS prediction can effectively improve user satisfaction on discovery results and reduce the time required for microservice instance discovery.