Enhancing Dependability of Cloud-Based IoT Services through Virtualization

As Internet of Things (IoT) technology moves forward, more and more IoT provided services are being pushed toward clouds. Since the operation of IoT services runs the risk of failures due to lossy communication links and error prone nature of physical objects, cloud providers (offering such services) should provide suitable platforms supporting two desired service dependability features -- i.e., reliability and availability. This issue has so far been addressed for specific application scenarios and often at the network layer. In this paper, we therefore aim at proposing a generic, model-based approach for enhancing these two important features at the application layer of cloud-based IoT systems. Following the principle of dependability by design, we build a framework based on the concept of virtualized IoT services, promising a service abstraction model to efficiently and simultaneously meet the dependability requirements of multiple cloud-based IoT applications. The proposed virtualization approach supports a variety of different dependability patterns and implements them according to the demands of the target application. We implemented the virtualization framework using the SicthSense cloud platform with satisfactory evaluation results on dependability metrics, such as maximum availability and the probability of failure on demand.