An Orchestration Based Cloud Auto-Healing Service Framework

As Cloud-deployed applications became more and more mainstream, continuously more complex services started to be deployed. The deployment involves multiple-layered systems and application itself. Where, failure of any point will disrupt the whole availability and auto-healing is one of the most indispensable features of the Cloud management. The micro-service model and Cloud management platform decouples application administrators from concrete infrastructure providers. This improves the efficiency of deliver and also brings up the challenge about how to separate the failure cases where the root causes lie in the infrastructure layer from others in the Cloud management platform and application itself. This separation will relief the burden of failure analysis for Cloud platform administrators and application owners. More importantly, the separation is the pre-condition for triggering the proper recover actions which can be different callbacks exposed by one infrastructure provider to another. In this paper, we propose an improved service framework with the auto-healing capability by adding a suit of customizable and flexible plugins into the common orchestration function of Cloud. The proposed auto-healing framework is suitable for any cloud applications to be health-cared, from Cloud-Native applications to Network Function Virtualization (NFV) applications. Our implementation is based on OpenStack and help out enterprise/telco use cases.