With the advent of 5G cellular networks, more stringent types of traffic, pertaining to applications like augmented reality, virtual reality, and online gaming, are being served nowadays. However, this comes with an increased energy consumption on both the user’s and network side, challenging this way the sustainability of cellular networks. Furthermore, the in-network computing aspect exacerbates things even further in that direction. 

Hence, it is very important to provide end-to-end sustainability, i.e., minimize the energy consumption in the network while maintaining performance guarantees, such as the maximum latency each flow should experience. This can be done, for example, depending on the traffic load in the network, and in order to keep the energy usage at low levels, the operator can decide to shut off certain network components, like User Plane Functions (UPFs) or edge clouds, and reassign the tasks to other entities. 

In this thesis, the focus will be on the core network. The aforementioned decisions will come up as solutions to optimization problems. To that end, the student will formulate optimization problems and solve them either analytically or using an optimization solver (e.g., Gurobi). The other part would be conducting realistic simulations and showing the improvements with our approach. 

- Basic understanding of 5G Core Networks and Mobile Edge Computing (MEC).

- Experience with mathematical formulation of optimization problems.

- Programming experience with Python and Gurobi.

Starting with 5G, assigning portions of network resources depending on the use case, where users running the same application/service would be getting resources from the same resource pool, a process known as network slicing, introduced a paradigm shift in how cellular networks operate in general. This brought significant advantages to both users and operators, by easing the resource allocation process and improving performance.

In 5G, users can be categorized, based on the service they are running, into three broad groups: eMBB, URLLC, and mMTC. There would be a separate Radio Access Networks (RAN) slice for each group. So, all eMBB users within the same cell would receive their resources from the same slice. URLLC service type is characterized by the most stringent traffic requirements. So, different amount of resources are needed to enable a satisfying user experience for different types of services. Hence, the problem of proper slice dimensioning arises as a very important issue that needs to be addressed.

Besides splitting RAN resources (slices), given that in-network computing will become an inevitable part of next generation of cellular networks, splitting computing resources would be very important. The computing resources come from edge clouds, which are collocated with base stations. Also, users have different channel conditions, which needs to be taken into account. For each admitted user, the traffic requirement would have to be satisfied.

Therefore, the operator would need to decide how to slice the RAN resources, and given the available computing resources, it would need to decide on the number of users of each service type to admit so that its overall revenue is maximized. The candidate needs to formulate an optimization problem. Then, different static and dynamic policies are to be analyzed in order to determine the best one, depending on the computational complexity introduced and the achievable performance in terms of the objective of the optimization problem.           

A good knowledge of any programming language is required.

The slice dimensioning for the three types of traffic in 5G(eMBB, URLLC, and mMTC) would be the focus of this thesis. Each user, depending on the type of traffic, is characterized by a weight. This could be e.g., the gain of the operator by serving a given user. We assume that the more stringent the traffic requirement is, the higher the gain for the operator is. In this way, mMTC users weight would be the lowest, whereas the URRLC’s the highest. Also, users have different channel conditions, which needs to be taken into account. For each admitted user, the traffic requirement would have to be satisfied. The problem would then reduce to deciding what sizes of network slices are to be allocated to each service type, and the corresponding number of users within the slice, so that the utility for the operator is maximized.

Three policies are to be analyzed. The first is motivated by the finite number of PhysicalResource Blocks (PRBs) each cell has, so that a brute-force solution is found. The computational complexity of this policy has to be obtained as well. The second, less complex policy, is the one in which resources are reserved beforehand (i.e., it is a static policy), based depending on the ratio between the weight and the resources needed for a user of the given service type. Finally, the third policy, which would be based on a heuristic, would decide onthefly on how to dimension RAN slice sizes. 

A good knowledge of any programming language is required.