PHM Technology for Memory Anomalies in Cloud Computing for IaaS

Abstract

The IaaS (Infrastructure as a Service) is one of the most popular services from todays cloud service providers, where the virtual machines (VM) are rented by users who can deploy any program they want in the VMs to make their own websites or use as their remote desktops. However, this poses a major challenge for cloud IaaS providers who cannot control the software programs that users develop, install or download on their rented VMs. Those programs may not be well developed with various bugs or even downloaded/installed together with virus, which often make damages to the VMs or infect the cloud platform. To keep the health of a cloud IaaS platform, it is very important to implement the PHM (Prognostics and Health Management) technology for detecting those software problems and self-healing them in an intelligent and timely way. This paper realized a novel PHM technology inspired by biological autonomic nervous system to deal with the memory anomalies of those programs running on the cloud IaaS platform. We first present an innovative autonomic computing technology called Bionic Autonomic Nervous System (BANS) to endow the cloud system with distinctive capabilities of perception, detection, reflection, and learning. Then, we propose a BANS-based Prognostics and Health Management (BPHM) technology to enable the cloud system self-dealing with various memory anomalies. AI-based failure prognostics, immediate self-healing, self-learning ability and self-improvement functions are implemented. Experimental results illustrate that the designed BPHM can automatically and intelligently deal with complex memory anomalies in a real cloud system for IaaS, to keep the system much more reliable and healthier.