XAI-Enabled Fine Granular Vertical Resources Autoscaler

Fine-granular management of cloud-native computing resources is one of the key features sought by cloud and edge operators. It consists in giving the exact amount of computing resources needed by a microservice to avoid resource over-provisioning, which is, by default, the adopted solution to prevent service degradation. Fine-granular resource management guarantees better computing resource usage, which is critical to reducing energy consumption and resource wastage (vital in edge computing). In this paper, we propose a novel Zero-touch management (ZSM) framework featuring a fine-granular computing resource scaler in a cloud-native environment. The proposed scaler algorithm uses Artificial Intelligence (AI)/Machine Learning (ML) models to predict microservice performances; if a service degradation is detected, then a root-cause analysis is conducted using eXplainable AI (XAI). Based on the XAI output, the proposed framework scales only the needed (exact amount) resources (i.e., CPU or memory) to overcome the service degradation. The proposed framework and resource scheduler have been implemented on top of a cloud-native platform based on the well-known Kubernetes tool. The obtained results clearly indicate that the proposed scheduler with lesser resources achieves the same service quality as the default scheduler of Kubernetes.