UC Irvine

Cloud computing is an essential part for current Internet of Things (IoT) applications. Despite cloud economic model's proven benefits, there are still some significant issues that cloud-based IoT applications may suffer from. One of the major issues is the disruption of IoT service caused by loss of connection between local resource-constrained devices and their cloud back-end which might be the result of cloud service outage or network link failure between local network to the cloud. Current solutions include using more powerful end IoT devices or adding IoT edge gateway in local network, but these solutions either significantly increase the cost or add a new single point of failure and break the simplicity in IoT device program design that cloud computing paradigm provides.

In this work, we argue for an edge service in local network as the backup of cloud service that is transparent to the IoT end devices. To this end, we propose an network-aware IoT edge service that includes 1)an architecture in which local IoT edge server cooperates with software-defined network (SDN) infrastructure to detect cloud/network failures based on IoT application's policy and perform failover methods following cloud-first manner 2)The network metric measurement methods that could be adopted in this architecture. We also present an implementation of our approach based on Floodlight SDN controller and Mininet emulator and conduct experiments that quantify performance for methods of cloud service failure detection and failover methods.