PERFORMANCE ANALYSIS OF EDGE COMPUTING FOR 5G AND INTERNET OF THINGS

This thesis aims to explore edge computing paradigm for Internet of Things services and applications in the 5G era. Edge computing, with its processing and storage capabilities near end-users, can become a reasonable alternative to cloud computing. User device with constrained processing and storage capabilities offloads complex task to edge nodes for computing, and then, edge nodes transmit the outcomes back to the user. Though various requirements need to be met while deploying edge computing in smart applications, this work focuses on the most demanding and critical ones, such as latency and system reliability. It is crucial to minimize latency and enhance the reliability of the system in time-critical services, such as smart healthcare or transportation. Current research implements a single user – multiple edge nodes model with Rayleigh and Nakagami-m fading channels. It is worth noting that Nakagami-m fading channels are widely used in the fifth generation and beyond systems. In order to comprehensively investigate the topic of edge computing for Internet of Things, we have developed two different types of schemes. Specifically, selection schemes, in which user device offloads task to one edge node, and combining schemes, in which user device offloads task to several edge nodes concurrently. Concerning selection-based schemes, edge computing with cache-aided relay and cache-free relay are considered. In these schemes, offloading nodes are chosen based on the best computing capability, channel gain between user and edge node, or channel gain between the user device and relay node. Numerical results demonstrate that edge computing model with cache-free relay, where the best channel gain between the relay and edge node is selected, outperforms other models in system performance. At the same time, time-division multiple access, frequency-division multiple access, and capacity achieving schemes are introduced for combining case. Thus, it can be summarized that edge computing model with capacity achieving scheme demonstrates the highest results among the others in terms of the reliability metric. Moreover, performance analysis shows that system efficiency can be affected by various parameters, such as transmit power, channel bandwidth, task size, latency threshold, number of edge nodes, and others. Numerical and simulation results are provided to validate analytical findings