Abstract

With the evolution of cellular networks, the number of smart connected devices have witnessed a tremendous increase to reach billions by 2020 as forecasted by Cisco, constituting what is known today as the Internet of Things (IoT). With such explosion of smart devices, novel services have evolved and invaded almost every aspect of our lives; from e-health to smart homes and smart factories, etc. Such services come with stringent QoS requirements. While the current network infrastructure (4G) is providing an acceptable QoE to the end users, it will be rendered obsolete when considering the critical QoS requirements of such new services. Hence, to deliver the seamless experience these services provide, MEC has emerged as a promising technology to offer the cloud capabilities at the edge of the network, and hence, meeting the low latency requirements of such services. Moreover, another QoS parameter that needs to be addressed is the ultra high reliability demanded by the IoT services. Therefore,5G has evolved as a promising technology supporting ultra Reliable Low Latency Communication (uRLLC) and other service categories. While integrating uRLLC with MEC would help in realizing such services, it would however raise some challenges for the network operator. Thus, in this thesis, we address some of these challenges. Specifically, in the second chapter, we address the problem of MEC Resource Provisioning and Workload Assignment (RPWA) in an IoT environment, with heterogeneous workloads demanding services with stringent latency requirements. We formulate the problem as an MIP with the objective to minimize the re-sources deployment cost. Due to the complexity of the problem, we will develop a decomposition approach (RPWA-D) to solve the problem and study through different simulations, the performance of our approach. In chapter 3, we consider both ultra high reliability and low latency requirements of different IoT services, and solve the Workload Assignment problem (WA) in an IoT environment. We formulate the problem as an MIP with the objective of maximizing the admitted workload to the network. After showing the complexity of the problem and the non scalability of the WA-MIP, we propose two different approaches; WA-D and WA-Tabu. The results show that WA-Tabu was the most efficient and scalable.