A Distributed NFV-Enabled Edge Cloud Architecture for ICN-Based Disaster Management Services

In disaster management services, the dynamic binding between roles and individuals for creating response teams across multiple organizations to act during a disaster recovery time period is an important task. Existing studies have shown that IP-based or traditional telephony solutions are not well-suited to deal with such group communication. Research has also shown the advantages of leveraging information centric networking (ICN) in providing essential communication in disaster management services. However, present studies use a centralized networking architecture for disaster management, in which disaster information is gathered and processed at a centralized management center before incident responses are made and warning messages are sent out. The centralized design can be inefficient in terms of scalability and communication. The reason is that when the network is very large (i.e., country level), the management for disaster services becomes very complicated, with a large number of organizations and offices. Disaster data are required to be transmitted over a long path before reaching the central management center. As a result, the transmission overhead and delay are high. Especially when the network is fragmented and network connectivity from a disaster-affected region to the central management center is disconnected, the service may be corrupted. In this paper, we designed and implemented a distributed edge cloud architecture based on ICN and network function virtualization (NFV) to address the above issues. In the proposed architecture, disaster management functions with predefined disaster templates were implemented at edge clouds closed to local regions to reduce the communication overhead and increase the service availability. The real implementation and performance evaluation showed that the proposed architecture achieves a significant improvement in terms of average bandwidth utilization, disaster notification delivery latency, routing convergence time, and successful request ratio compared to the existing approaches.