Motivation:

Deutsche Telekom is operating and constantly developing and improving it’s own cloud to operate internet and telephony services (Telco applications). The Kubernetes-based cloud and the Telco applications are combined to form a TaaP - Telco as a Platform. The TaaP are thousands of servers and hundreds of applications. The energy efficiency of the TaaP is a key success criterion for optimizing costs, energy consumption, and carbon emissions. Hence, a concept of Full Stack Energy Management is established. The focus is to jointly optimize hardware, software, and services towards energy efficiency without affecting service availability and robustness. 

Problem & Challenge:

In the Telco industry, so far, HW redundancy has been the baseline for service robustness and resilience. The introduction of virtualization and containerization concepts resulted in an additional redundancy level above the hardware. Classical redundancy models no longer apply to this multi-layer Software redundancy. Moreover, so far, there is no mathematical model that captures the service availability for these new architectures. For example, a Telco-Service can be provided via 3 data centers with 50 servers each, forming 1 cluster hosting 500 Kubernetes pods. There is a mix of data center redundancy, hardware redundancy in the data centers, and Kubernetes worker node and pod redundancy.

Specific Problem Formulation:

On a TaaP there are multiple layers of redundancy in Hardware and Software. On the one hand, there are multiple site deployments, where each site has multiple hundreds of servers. On the other hand, on each site, each server has multiple redundant hardware parts, like the power supply. Moreover, a Kubernetes Cluster, which is homed on one site, hosts multiple microservices, each with a different redundancy concept like active/passive, n+1, n+m, etc. This setup of mixed HW and SW redundancy causes inefficiency and is not easy to calculate or simulate in terms of overall service availability, network, site, redundancy, and energy consumption. 

Solution Approach:

There are multiple different parameters in HW and SW that impact the service availability and energy consumption. Firstly, a comprehensive list of these parameters is required, including modeling of dependencies. Secondly, a mathematical model needs to be set up that considers all of these parameters in "one equation". Thirdly, a graphical simulation should be set up to demonstrate the dependencies and results.

Expected Outcome:

A simulation and mathematical model should be developed that considers software and hardware redundancy across multiple sites and SW layers to calculate the network-wide service availability. This is key in order to further optimize the HW and SW footprint and improve sustainability. Moreover, the model should allow the optimization of the following parameters: least required HW based on predefined service availability, least energy consumption, and best redundancy.

Working on the live network or setting up a lab deployment is out of scope of this thesis. The focus of this thesis are the modeling and simulation of the deployment.

Voraussetzungen

Within current 6G research, so-called in-X Subnetworks (SNs) are envisioned to constitute a key technology for providing ubiquitous network connectivity. Possible deployment scenarios for these SNs include, e.g., industrial environments, vehicles, or drones to facilitate communication between users in geographically confined areas with high data rates, ultra-low latencies, and high reliability. The use-cases that are envisioned for in-X SNs are, e.g., intra-vehicle sensor-actuator communication, robot control in industrial environments, or in-body networks for health monitoring. SNs can thereby be operated in three different modes: In the standalone mode, the SN runs fully autonomous, while in the semi-autonomous and connected mode, the SN can receive control information from a so-called 6G umbrella network, and the SN Access Point (AP) can act as a gateway to the umbrella network.

Regardless of the operation mode, and despite SNs representing a new type of network, SNs still require frequency bands to provide their services. Therefore, adequate frequency planning mechanisms must be developed to mitigate interference among SNs, especially for SNs enabling high reliability communication services. Due to the deployment of SNs within autonomous vehicles, Unmanned Aerial Vehicles (UAVs), or robots, many of the SNs will be mobile. The mobility of SNs thus makes the problem of frequency planning for SNs a dynamic task. Since frequency spectrum is a limited resource, resource allocation mechanisms must focus on effective spectrum usage while ensuring a certain communication quality.

While there exists related work addressing this problem for mobile SNs, i.e., for varying interference scenarios, and optimizing resource allocation at different time instances, SN reconfigurations, i.e., changing a SN’s allocated frequency spectrum from one time step to the other, have not yet been considered. Due to the signaling overhead, a SN reconfiguration can, however, have significant impact on the service availability within a SN. Thus, in [1], the problem of dynamic frequency subband allocation to mobile in-X SNs with the objectives of minimizing SN reconfigurations and bandwidth usage while guaranteeing interference-free operation of all SNs has been studied and solved using heuristic algorithms.

To complement this work, this thesis should investigate different solution approaches based on Machine Learning (ML) to solve the problem presented in [1]. Specifically, in the case of a research internship, the student should implement a Graph Neural Network (GNN) to solve the optimization problem. In the case of a master thesis, it is also possible to change the system model and problem setting. Depending on the quality of the work, the results of the thesis can lead to a scientific publication.

[1] V. T. Haider, R. Pries, W. Kellerer, and F. Mehmeti, “Dynamic frequency planning for autonomous mobile 6G in-X subnetworks,” in NOMS 2025-2025 IEEE/IFIP Network Operations and Management Symposium, To Appear, 2025.