Workload Prediction and its Impact on QoS of Microservices Architecture Based Applications

Microservices is an architectural style used for developing modular and scalable applications. The distributed and de centralised architectural property of microservices applications makes it difficult to manage unpredictable workload variations and sometimes lead to degraded QoS of the applications. Accurate workload prediction is important for efficient resource allocation and maintaining high QoS of the MSA based applications. This paper proposes an LSTM-RNN based model to predict the workload of each microservice instance and after that an analysis is done to identify the effects of proactive workload management on the QoS metrics such as response time, throughput, and CPU utilisation. The results show that the workload prediction accuracy significantly improves the latency, resource utilisation and performance of microservices applications.