Abstract

The Internet of Things (IoT) enables objects to interact and cooperate with each other for reaching common objectives. It is very useful in large-scale disaster management systems where humans are likely to fail when they attempt to perform search and rescue operations in high-risk sites. IoT can indeed play a critical role in all phases of large-scale disasters (i.e. preparedness, relief, and recovery). Network softwarization aims at designing, architecting, deploying, and managing network components primarily based on software programmability properties. It relies on key technologies, such as cloud computing, Network Functions Virtualization (NFV), and Software Defined Networking (SDN). The key benefits are agility and cost efficiency. This thesis proposes softwarization approaches to tackle the key challenges related to large-scale IoT based disaster management systems.
A first challenge faced by large-scale IoT disaster management systems is the dynamic formation of an optimal coalition of IoT devices for the tasks at hand. Meeting this challenge is critical for cost efficiency. A second challenge is an interoperability. IoT environments remain highly heterogeneous. However, the IoT devices need to interact. Yet another challenge is Quality of Service (QoS). Disaster management applications are known to be very QoS sensitive, especially when it comes to delay.
To tackle the first challenge, we propose a cloud-based architecture that enables the formation of efficient coalitions of IoT devices for search and rescue tasks. The proposed architecture enables the publication and discovery of IoT devices belonging to different cloud providers. It also comes with a coalition formation algorithm. For the second challenge, we propose an NFV and SDN based - architecture for on-the-fly IoT gateway provisioning. The gateway functions are provisioned as Virtual Network Functions (VNFs) that are chained on-the-fly in the IoT domain using SDN. When it comes to the third challenge, we rely on fog computing to meet the QoS and propose algorithms that provision IoT applications components in hybrid NFV based - cloud/fogs. Both stationary and mobile fog nodes are considered. In the case of mobile fog nodes, a Tabu Search-based heuristic is proposed. It finds a near-optimal solution and we numerically show that it is faster than the Integer Linear Programming (ILP) solution by several orders of magnitude.