Future wireless networks will be characterized by heterogeneous traffic requirements. Examples can be low-latency or minimum-througput requirements. Therefore, the network has to adjust to different needs. Usually, users with low-latency requirements have to deliver their demand within a specific time frame, i.e., before a deadline, and they coexist with throughput oriented users. In addition, mobile devices have a limited-power budget and therefore, a power-efficient scheduling scheme is required by the network. In this work, we cast a stochastic network optimization problem for minimizing the packet drop rate while guaranteeing a minimum throughput and taking into account the limited-power capabilities of the users. We apply tools from Lyapunov optimization theory in order to provide an algorithm, named Dynamic Power Control (DPC) algorithm, that solves the formulated problem in real time. It is proved that the DPC algorithm gives a solution arbitrarily close to the optimal one. Simulation results show that our algorithm outperforms the baseline Largest-Debt-First (LDF) algorithm for short deadlines and multiple users.

5G and beyond networks are driven by the vision of the Internet of Things and mission-critical communications that have time-sensitive applications requiring low latency. These new, related to the previous network generations, requirements are challenging and they have a technical impact on the design of communication networks and classical networks cannot support such requirements. To fulfill those requirements, we need to design technologies and techniques across all the layers of the network. Moreover, time-sensitive applications, e.g., a cyber-physical system with sensors, may require fresh data. The traditional notion of latency is not sufficient to characterize the freshness of the data. However, a new metric has been proposed that captures the freshness of the data and it is called Age of Information (AoI). AoI takes into account not only the packet delivery delay but also the times between the generation of the information.

Besides the time-critical applications, communication networks have to support the classical Mobile Broadband (MBB) users that usually have throughput requirements. Time-critical applications together with MBB users create a network with heterogeneous traffic. Therefore, 5G and beyond networks will be characterized also by the heterogeneity of the traffic. The goal of thesis is to study the performance of networks with heterogeneous traffic including time-sensitive applications and provide solutions for optimizing it.

In the first paper, we provide a dynamic algorithm with low complexity that schedules users with heterogeneous traffic. In particular, we consider a wireless network with two sets of users; i) users with deadline-constrained traffic, ii) users with minimum throughput requirements. In addition, the users have a limited power budget. This work aims at the minimization of the packet drop rate while achieving the minimum throughput requirements, and average power below a threshold. In order to achieve our goal, we cast a stochastic optimization problem and use the Lyapunov optimization framework to solve it. We propose a dynamic power control algorithm that is proven to provide a solution arbitrarily close to the optimal one.

The second paper investigates the impact of sampling cost in a wireless network with AoI-oriented users that sample and transmit their information over an erasure wireless channel to a receiver. The users can decide to sample and transmit a new status update or to send an old one. Our goal is to study the impact of storing non-fresh packets on the total system cost when the sampling and transmission costs are not negligible. We formulate a stochastic optimization problem for minimizing the average total cost while providing fresh enough data to the receiver. Three scheduling policies are provided, as well as their performances through simulation results. We observe that having the option to send an old packet can significantly improve the total system cost.

In the third and fourth papers, we consider a wireless network consisting of an AoI-oriented user and a deadline-constrained user. More specifically, in the fourth paper, we study the interplay between packet drop rate and the average AoI of users in a multi-access channel. We provide analytical expressions for the average AoI by modeling the evolution of the AoI as a Discrete- Time Markov Chain (DTMC). Furthermore, we model the remaining time of the packet with the deadline as a DTMC, and by utilizing its steady-state distribution, we provide the analytical expression of the packet drop rate. Inspiring by the third paper, the fourth paper considers the AoI minimization with timely-throughput constraints. Both time-correlated and independent identically distributed channel models are considered. The problem is formulated as a CMDP and is relaxed by utilizing tools from Lyapunov optimization theory. The relaxed problem is a finite horizon MDP problem which is solved by applying backward dynamic programming.

In the fifth paper, we study the performance of a network that consists of Virtual Network Functions (VNF) and Multi-access Edge Computing (MEC) servers. The tasks that are transmitted by a user to the base station have to be processed by two VNFs. The VNFs are located both at the MEC and core server. A MEC server is co-located with the base station, and another is close to the base station. We aim to study the impact of offloading traffic to the MEC acting as a helper to the end-to-end delay, throughput, and task drop rate. We provide a methodology for deriving analytical expressions for the metrics of our interest. This methodology can be applied to larger and more general systems. Simulation results show the interplay between those metrics under different network set-ups.

In the sixth paper, we consider an Unmanned Aerial Vehicle (UAV) that flies over multiple mobile areas for serving users within a specific time horizon. Our goal is to maximize the number of served users. We formulate an optimization problem for maximizing the served users under a time constraint, i.e., a deadline. The problem is proved to be NP-hard. We provide a greedy low complexity algorithm that can solve the problem in real-time. Simulation results show that our algorithm approximates well the optimal solution, and it is efficient regarding the running time even for large networks.

Future mobile networks supporting Internet of Things are expected to provide both high throughput and low latency to user-specific services. One way to overcome this challenge is to adopt Network Function Virtualization (NFV) and Multi-access Edge Computing (MEC). Besides latency constraints, these services may have strict function chaining requirements. The distribution of network functions over different hosts and more flexible routing caused by service function chaining raise new challenges for end-to-end performance analysis. In this paper, as a first step, we analyze an end-to-end communication system that consists of both MEC servers and a server at the core network hosting different types of virtual network functions. We develop a queueing model for the performance analysis of the system consisting of both processing and transmission flows. We propose a method in order to derive analytical expressions of the performance metrics of interest, i.e., end-to-end delay, system throughput, task drop rate. Then, we show how to apply the similar method to a larger system and derive a stochastic model for such systems. We observe that the simulation and analytical results are very close. By evaluating the system under different scenarios, we provide insights for the decision making on traffic flow control and its impact on critical performance metrics.

In this letter, we consider a UAV flying over multiple locations and serves as many users as possible within a given time duration. We study the problem of optimal trajectory design, which we formulate as a mixed-integer linear program. For large instances of the problem where the options for trajectories become prohibitively many, we establish a connection to the orienteering problem, and propose a corresponding greedy algorithm. Simulation results show that the proposed algorithm is fast and yields solutions close to the optimal ones. The proposed algorithm can be used for trajectory planning in content caching or tactical field operations.

Future communication systems will be characterized by heterogeneous traffic and requirements. Time critical applications like cyberphysical systems, augmented and virtual reality, raise the need for a low-latency based network. At the same time, conventional devices requiring for high throughput will co-exist with time critical applications. Besides the new technologies, new scheduling and optimization techniques are needed to face these challenges. 

In this thesis, we investigate the issues arising from the deployment of these technologies. In Paper I, we explore the benefits of dynamic Transmission Time Interval (TTI) selection in a heterogeneous network environment. We consider packets with deadlines and we optimize jointly the TTI length and the channel allocation. After proving the NP-hardness of the problem, we propose a greedy algorithm taking decisions in polynomial time. The first work opens new questions regarding the deadline-constrained traffic such as how the minimum average drop rate can be achieved. In Paper II, we consider power-limited devices with deadline-constrained traffic. Lyapunov optimization methods are explored to solve the problem with time average objective and constraints. We develop a dynamic, polynomial time, algorithm that finds an approximation of the dropping rate minimization problem under average power constraints. 

Besides the new techniques, future communication systems will require the development of new technologies for a more exible and elastic network. Multi-access Edge Computing (MEC) and Virtual Network Function (VNF) technologies are considered two of the key technologies for next generation networks. In Paper III, we analyze the performance of a network that hosts VNF and consists of MEC servers and servers at the core. As a first step, we consider a simple end-to-end communication system and provide analytical expressions for the end-to-end delay and system throughput by applying tools from queueing theory. Based on the first step, we provide the methodology for analyzing scaled-up systems with arbitrary number of servers. Simulation results show that our analytical model performs well. Furthermore, this work provides insights for the design and performance optimization of such systems such as optimal ow control and resource allocation.