A novel approach for enhancing reliability of microservices applications under dynamic workloads

A novel approach for enhancing reliability of microservices applications under dynamic workloads

연구 분야: Software Development

학회: Cluster Computing

The increasing need for independently developed and deployed web-based applications has made microservices architecture popular in the last decade. Its benefits have incentivized large IT companies like Netflix, Uber, and Twitter to adopt it. However, the distributed and loosely coupled nature of microservices based applications introduces reliability challenges under dynamic workloads. The failure of a microservice in the presence of dynamic workloads makes its dependent microservices unresponsive causing cascading failures and communication breakdowns resulting in reduced reliability. Effective workload handling and robust microservices communication mechanisms are essential to sustaining the reliability of microservices applications. This paper proposes a novel approach for the reliability improvement of microservice applications by focusing on these two issues. The proposed approach consists of four phases, the first phase assesses the reliability of microservices applications under varying workloads in the absence of the proposed approach and these results are used later for performing empirical analysis. The second and third phases of the proposed approach have focused on effective workload handling by predicting the microservices instances based on the forecasted workload. The fourth phase reduces the potential points of failure associated with microservices communication by placing interdependent microservices on the same node. To assess the performance of the proposed approach, an empirical investigation is performed on a case study application by employing synthetic and realistic workloads. The experimental results have shown that the proposed framework has significantly improved the reliability of the microservices application.