Understanding Data Analysis in an End-to-End IoT System
Abstract
The Internet of Things (IoT) is known as the concept of connectingeveryday physical devices to the Internet. It is natural to assume that thepopularity and development within this field will increase in the followingyears. This means that more and more things will be able to communicateover the Internet. In the process of developing IoT, an important part is tobuild reliable and scalable networks, and understanding where data shouldbe processed concerning power consumption and costs of transferringdata in various regions of the network.
The task of the thesis will be to access data in a complete prototypeof an IoT network, and both collect and analyse the data. The goal isto study different alternatives for a typical IoT system and provide anoverview of current state-of-the-art technologies, products and standardsthat can be used in such a setting. Data can be generated by using andcomparing different sensors connected to end nodes in the network. Acomplete network of both microcontrollers and single-board computerswill be built and explained in this thesis. The network will from now onbe referred to as testbed.
Microcontrollers as end nodes in an IoT network will be the central elementtested in this thesis. The primary focus is to establish a connectionbetween two devices, A and B, and form a network between these thatcan transport data efficiently. A central point of discussion will be to findtransfer protocols and technologies that can be used in such a system. Itwill be discussed the advantage and disadvantage of sending raw data,rather than doing the computation in the end nodes. The main focus willbe on optimal throughput in the network. A deep understanding of thebenefits of processing data in the end nodes, concerning power, costs andtime is needed to achieve this.
Results from this work include graphs and discussions explaining in whichcase the different transport protocols suggested are preferred, from testsdone in the testbed. These show that different protocols are suited fordifferent usage and that one of the tested possibilities more stable thanthe other in the tests presented. Both protocols registered their highestmeasured goodput at approximately 600 bytes/second. Being a quite slowtransfer rate, this opened up for another discussion about the possibleuse cases for future Bluetooth Low Energy (BLE)-based IoT applications.
Keywords: Optimizing payload sizes, fragmentation, maximizing through-put, power usage.