Master Thesis Defense: Md. Nazmul Alam
Speaker: Md. Nazmul Alam
Supervisor: Dr. R. Glitho
Examining Committee: Drs. Y.-G. Gueheneuc,E. Shihab, J. Rilling (Chair)
Title: An Architecture for an Infrastructure as a Service for Internet of Things
Date: Wednesday, June 20, 2018
Place: EV 11.119
Internet of things (IoT) refers to things such as sensors and actuators interacting with each other to reach common goals. It enables multiple applications in sectors ranging from agriculture to health. Nowadays, applications and IoT infrastructure are tightly coupled and this may lead to the deployment of redundant IoT infrastructures, thus, cost inefficiency.
Cloud computing can help in tackling the problem. It is a paradigm to quickly provision configured resources (computing, network, memory) on demand for cost efficiency. It has three layers, the infrastructure as a service (IaaS), the platform as a service (PaaS) and the software as a service (SaaS). Through the IaaS, configured hardware resources (CPU, storage, etc.) are provisioned on demand. However, designing and implementing an IoT IaaS architecture for the provisioning of IoT resource on demand remains very challenging. An example of a challenge is using appropriate publishing and discovery mechanism suitable for IoT devices. Orchestrating virtualized IoT device over several physical IoT devices is an example of another challenge that should be addressed.
The main contribution of this thesis is twofold. First, a novel IoT IaaS architecture is proposed where IoT devices can be provisioned as a configured infrastructure resource on demand via node level virtualization. Second, the architecture is prototyped and evaluated using real-life sensors that support node-level virtualization. It is also shown how node level virtualization can achieve resource efficiency in contrast to middleware based solutions. The essential architectural features, such as publication, discovery, and orchestration are identified and proposed. Two sets of a high-level interface are also introduced. A low-level uniform interface is suggested to decouple the IoT devices from the applications by allowing the applications to access the heterogeneous devices in a uniform way. In addition, a cloud management interface is proposed to expose the IoT IaaS to the cloud consumers (e,g - the PaaS, the application, etc.) and allow them to provision the IoT resources.