Proactive Autoscaling for Cloud-Native Applications using Machine Learning
Nicolas Marie-Magdelaine, Toufik Ahmed
Abstract
Cloud computing and cloud-native applications have become standards for new developments in most organizations. Many companies are moving their workload toward microservices and profit from the cloud-native paradigm. However, there are still many challenges to overcome to optimize the Quality of Service using autoscaling and resource dimensioning. Autoscaling is defined as the capacity of cloud infrastructure and applications to resize themselves by adjusting the resource pool. Simple autoscaling can be achieved by a feedback-loop based on current workload resource utilization. Called reactive autoscaling, this technique may introduce inconsistency between workload and resource utilization. In this paper, we propose a proactive autoscaling framework using a learning-based forecast model to dynamically adjust the resource pool, horizontally (number of replicas) and vertically (resources pool). This framework uses a proactive autoscaling algorithm based on Long Short-Term Memory (LSTM) to improve the end-to-end latency for cloud-native applications. We developed a proof of concept to demonstrate this framework and the effectiveness of our proactive algorithm.