Seminar

Driven by the requirements of innovative applications and services, the past few years have produced new technologies for networked or distributed systems. In the Internet of Things, a large number of devices and everyday objects equipped with sensors and actuators are connected to the internet and communicate through mostly wireless communication technologies. In one of its sub-domains, the industrial Internet of Things (IIoT, Industry 4.0), machines, tools, transport equipment, etc. are networked.
Virtualisation and "software-defined" systems (e.g. Software Defined Networking (SDN)), increase the flexiblity and efficiency of distributed systems through dynamic adaptation and scaling.
Driven by the popularity of the Bitcoin system, distributed ledger technologies and related concepts such as smart contracts have been developed, which are not only the foundations of electronic currencies, but can also support any application in which a consensus between different parties must be reached and documented. 
Another focus has been the reduction of latency in networked and distributed systems, e.g. by using nearby edge and fog computing resources in addition to the remote cloud, or by using optimised communication protocols to., for instance, rapidly connect client and server. 
In addition to stationary networks, mobile (5G) communication technologies and systems have developed rapidly. For example, the COVID tracing application uses mobile devices to track contacts. This method, known as crowdsensing, can be used more generally to collect large amounts of geographically distributed sensor data. 

This seminar will discuss a wide range of current technologies, protocols and standards that enable the above networked and distributed applications and services. 

The seminar is organized in the style of a scientific conference. Following the submission of a written paper on the assigned topic, students write reviews for other seminar papers and participate in a final presentation session where they present their work and discuss the work of others. Attendance at the kick-off and final presentation session is mandatory.

Flexible, scalable, and efficient networks

1) QUIC -The Fast TCP Alternative

Supervisor: Michael Schramm

For large web companies and Content Distribution Networks the performance and security of web protocols is of big concern. For this reason, Google proposed the QUIC protocol. All of Google’s mobile applications support QUIC and by October 2020, more than 75% of the Meta internet traffic uses QUIC, making it to the standard internet protocol at Meta¹. How does this upcoming protocol ensure security? How is the higher performance achieved and how much higher is it?

The objective of this seminar topic is to explain the QUIC protocol and compare it with the TCP protocol variants. It should also review scientific papers with regard to performance measurements and experiences with the QUIC protocol.

2) Routing Protocols in Vehicular Ad-Hoc Networks (VANETs)

Supervisor: Michael Matthé

Vehicular Ad-Hoc Networks (VANETs) is an important field of research for intelligent transportation systems. It has the potential to improve, among others, road safety and traffic monitoring. Due to the high mobility of vehicles inside a VANET, the topology of the network is subject to frequent changes. Therefore, the selection of routing protocols is essential to ensure the desired operation of VANETs and the connected applications.

The objective of this seminar topic is to review the state of the art in VANET routing protocols. This may include looking at how the routing protocols differ from more traditional mobile ad-hoc networks and novel approaches such as reinforcement learning-based routing protocols.

3) Low-Latency Internet

Supervisor: Robin Laidig

Low latency is crucial for many networked applications and often improves performance and user experience. Prominent examples are online games, where high latencies can even make them unplayable; At Amazon, even a small increase in latencies on the online shopping platform leads to lower sales, with a loss amounting to almost 1 billion dollars. But why are latencies in the internet so high and how can we improve it?

The objective of this seminar topic is to review computer science research that proposes solutions to overcome high latencies in the internet.

P2P networks

4) Peer2Peer Energy Trading

Supervisor: Sokol Makolli

More and more energy consumers are starting to also produce and sell their own energy (34,7% increase from 2021 to 2022 in Germany). Against this backdrop, the opportunity becomes apparent to cut out the middleman and enable these so-called prosumers to trade energy freely among themselves. This is called Peer2Peer energy trading. While such an implementation may sound trivial at first glance, it is important to keep in mind that the electricity grid is quite special and fragile. A stable grid requires a constant balance between supply and demand. Thus, a peer2peer trading system must ensure that each feeding-in always finds its exact match with the energy demand.

The objective of this seminar topic is to introduce currently researched approaches of Peer2Peer energy systems and to systematically compare them.

Applied edge computing

5) Offloading at the Edge: A Review of Edge Computing Use Cases

Supervisor: Melanie Heck

The idea of edge computing is simple: Instead of using centralized cloud resources, applications offload tasks to devices at the edge of the network such as smartphones, PCs, or small servers attached to cell towers. This reduces latency and makes better use of idle end-user devices. One question, however, often remains unanswered: What applications are particularly suitable for offloading in the edge?

This seminar topic explores potential real-world use cases, with the objective to provide a systematic overview.

6) Unlimited Power: Code Offloading in the Internet of Things

Supervisor: Melanie Heck

In the Internet of Things (IoT), a growing number of devices and objects equipped with sensors are connected and collect a constant stream of huge volumes of data. Processing the data on centralized cloud servers puts pressure on the network and causes delays. At the same time, time-critical applications such as virtual and augmented reality require real-time processing. Edge computing therefore offloads the computation to nearby devices and decentralized servers at the edge of the network. In order for offloading to be effective, important decisions have to be made regarding what, when, where, and how to offload code.

The objective of this seminar topic is to compare and critically evaluate approaches that coordinate and orchestrate code offloading from resource-constrained IoT devices to nearby edge servers.

Latency and energy-aware offloading

7) Proactive Data Placement in Mobile Ad-Hoc Networks

Supervisor: Melanie Heck

Offloading computationally intensive tasks can speed up the computation. However, tasks like machine learning require transferring large amounts of input data. This makes offloading more time-consuming. To minimize delays, data files can be proactively placed on potential future resource providers. In mobile ad-hoc networks, resource providers that are candidates for future offloading tasks can be identified by analyzing movement patterns. The characteristics of ad-hoc computing environments necessitate non-trivial data and task placement strategies.

The objective of this seminar topic is to review approaches for proactive data placement in mobile ad-hoc networks.

8) Offloading for Mobile Devices – Trading Computation for Communication

Supervisor: Johannes Kässinger

Computationally intensive mobile applications consume a lot of energy. One possible solution is to offload parts of the computation to remote cloud or edge servers. But in doing so, we trade computation energy for communication energy. The challenge is to have an architecture where the saved computational energy is not less than the energy that is now required for communication. Additionally, real-time applications like augmented reality require receiving the results in time.

This seminar topic surveys different offloading architectures, including edge computing, which has a lower delay due to the closer positioning of resource providers to the offloading device. The objective is to discuss all methods with respect to real-time applications considering delay, local computation, communication, and compression.

9) Balancing Energy and Latency for Offloading in Edge Computing

Supervisor: Melanie Heck

With the widespread availability of cloud services, more and more computing tasks are being outsourced from end-user devices to remote servers. Code offloading not only drastically reduces computation time, but also conserves energy on the offloading device. This is particularly beneficial for mobile devices with limited battery life. Edge computing reduces transmission latencies when offloading tasks by placing computational resources at the edge of a network, such as cell towers. The selection of a suitable resource provider for fast and energy efficient task offloading depends heavily on the available bandwidth, task execution time and data transfer time.

This seminar topic reviews energy and latency aware offloading approaches. The objective is to evaluate the approaches with regard to how the authors estimate the available bandwidth and the time required to transfer and execute a task.

10) Energy-Aware Computation Offloading in Edge Computing Environments

Supervisor: Melanie Heck

Smartphones and wearables have limited computational resources. To perform computationally intensive tasks such as machine learning, the computation can be offloaded to remote resource providers. As an alternative to the cloud, edge computing offloads the computation to nearby base stations or end-user devices. If a suitable resource provider is chosen, this not only prevents the offloading device from quickly running out of battery, but can also reduce the overall energy consumption in the edge computing environment.

The objective of this seminar topic is to review and systematically present offloading approaches for edge computing environments that specifically target energy savings – either at the offloading device, or for the entire distributed system including both resource consumers and providers.

Trust, privacy and security

11) Privacy Mechanisms for Computation Offloading at the Edge

Supervisor: Melanie Heck

Edge computing offloads computation to nearby devices that are often unreliable and untrusted. This introduces new threat vectors. To protect sensitive data and preserve user privacy, security and privacy mechanisms need to be deployed at multiple points in the network.

This seminar topic reviews mechanisms for protecting user privacy in edge computing environments, with the objective to systematically categorize them.

12) Measuring Trust in Edge Computing Environments

Supervisor: Melanie Heck

Multi-access edge computing offers fast, cost-effective, and bandwidth-efficient data processing close to its source. However, as data and computation are offloaded to unreliable devices at the edge, concerns for the security of sensitive data and the reliable execution of critical tasks are bigger than ever. The key challenge now is to determine trust levels for each device and allocate particularly sensitive tasks to the most reliable and trusted resource providers.

The objective of this seminar topic is to review trust management models and approaches that assess trust for both applications and resources in dynamic edge computing environments.

13) TOR - The Onion Router

Supervisor: Lukas Epple

Tor is a popular anonymity network that allows users to browse the internet without revealing their true identity, location, or activity. By routing internet traffic through a series of encrypted nodes or relays, the source of the traffic is obscured and user privacy is preserved.

This seminar topic explores how Tor works internally and how it achieves privacy. Additionally, the potential risks and limitations of Tor, as well as the impact it has on internet privacy and security are reviewed. The objective is to provide a comprehensive overview of Tor and its technical underpinnings.

Melanie Heck

The seminar is organized in the style of a scientific conference. Following the submission of a written paper on the assigned topic, students write reviews for other seminar papers and participate in a final presentation session where they present their work and discuss the work of others. Attendance at the kick-off and final presentation session is mandatory.

Time-Sensitive Networks (TSN)

1) How to Schedule Time-Triggered Data Flows Using TSN Technologies via Wireless Connections

Supervisor: Heiko Geppert

The IEEE has created a series of standards to enable deterministic real-time ethernet networks, known as Time-Sensitive Networks. For a network to be deterministic, the network delays must be bounded and known. When considering IEEE 802.11 wireless networks, the system becomes gradually more complex. Challenges including increased latency, reduced bandwidth, and collisions make it non-trivial to integrate wireless connections into TSN networks.

The objective of this seminar topic is to provide an overview of the problems and current solutions when using TSN with wireless networks. The focus should be on how to integrate the additional properties and issues of wireless connections into the already well known and researched abstract TSN models.

2) Online Scheduling Heuristics for Time-Triggered Real-Time Data Flows

Supervisor: Heiko Geppert

Modern Industrial IoT applications, such as automated industrial plants, require real-time communication between machines. Otherwise, a delayed message could lead to crashes resulting in human harm or financial loss. The IEEE 802.1 TSN standards define how ethernet networks can gain real-time properties. Features like the Time-Aware Shaper (TAS) enable deterministic schedules for time-triggered flows. However, the standards do not specify how to compute these schedules. Scheduling strategies developed by researchers range from offline algorithms yielding optimal solutions to online solutions returning valid schedules in a very short time.

The objective of this seminar topic is to review and present very fast, online capable scheduling algorithms for time-triggered traffic in TSN networks. Thereby, different feature sets, e.g., multi cast, and the scalability to large problem instances need to be considered.

3) Deterministic Latency Guarantees for Event-Based Network Traffic

Supervisor: Robin Laidig

With the Internet of Things and Industry 4.0, an increasing number of devices are connected that vary greatly in functionality and importance. Therefore, Quality of Service (QoS) models are essential to separate and prioritize their network traffic. Real-time network traffic that must arrive at its destination before a certain deadline has to be prioritized, otherwise catastrophic failures can occur. Scheduling algorithms with deterministic latency bounds guarantee delivery before a strict deadline. However, many of these algorithms are designed for periodic, time-triggered network traffic that can be precisely predicted. In reality, event-based (sporadic) network traffic that can emerge randomly is often present.

The objective of this seminar topic is to review computer science research that proposes network scheduling algorithms for event-based network traffic. Further, the seminar work shall point out future research suggestions.  

P2P networks

4) IPFS - Unstoppable Filesharing

Supervisor: Lukas Epple

IPFS is a decentralized network that allows to store and share files in a peer-to-peer (P2P) architecture. By distributing files across multiple nodes in the network, IPFS provides a more resilient and efficient way of sharing content. This approach eliminates the need for a central server and reduces the risk of censorship or network failures.

The objective of this seminar paper is to review how the IPFS works internally, what concepts of distributed systems are used, and what limitations the IPFS has.

Federated learning

5) Federated Learning With GNN

Supervisor: Michael Schramm

Graph neural networks (GNN) have proven their usefulness for learning on graph data in many different fields like recommender systems, traffic, finance, chemistry, etc. Unlike images, graph data is often not simply publicly available, but distributed in data silos. Data of institutions and companies often exists in isolated form because of privacy and commercial concerns, which is a problem for centralized GNN. Companies are in the predicament of not having a sufficient amount of data on their own to train a model, but would still like to be able to use GNNs. An example are pharmaceutical research institutions which would like to utilize GNN, but their data is limited and often confidential.

Federated Learning is a machine learning setting where many clients train together without making the data itself accessible to the collaborators. This enables institutions with smaller datasets to gain insights they could not get from their own data alone. Federated Learning can also help in areas where privacy is restricted, such as medical data.

This seminar topic reviews scientific publications about Federated Learning on GNN. The objective is to describe different types of Federated Learning in terms of data distribution, shared data and system model. Furthermore, applications of Federated Learning on GNN should be discussed.

Offloading and edge computing

6) Traffic Control with Vehicular Edge Computing

Supervisor: Michael Matthé

Edge computing allows applications to utilize local data storage and processing to provide low delay and high bandwidth. Vehicular networks can leverage these advantages by offloading computation-intensive and latency-sensitive tasks (Vehicular Edge Computing).

This seminar topic examines vehicular traffic management, one of the application areas of vehicular edge computing. Vehicular Traffic Management approaches can benefit from vehicular edge computing to provide increased traffic efficiency, as well as improvements in road safety. Some of the challenges that need to be considered to employ vehicular edge computing effectively are the high mobility of devices (vehicles), dynamic channel environment, and resource management of the local resources.

7) Synchronization in Online Games – What’s Good for Games is Good for Simulations?!

Supervisor: Johannes Kässinger

Computationally intensive mobile applications like simulations consume a lot of energy. One possible solution is to offload computation to remote cloud or edge servers. A key challenge is to synchronize the results when multiple offloading resources are used. Additionally, real-time applications like augmented reality must receive a valid result in time. Similar to online games, methods to compensate for delayed or lost updates are needed.

The objective of this seminar topic is to survey existing synchronization and compensation methods in online games. In the second part of the paper, assumptions shall be made on how these methods can be used in a mobile simulation that relies on results from (multiple) remote servers.

Smart grid

8) Energy System Modeling for Africa

Supervisor: Sonja Klingert

In step with the energy turnaround, many African countries need the diffusion of energy, mostly electricity, to supply a higher share of the population. For this expansion, however, energy system models are needed that capture the specific characteristics of the current energy systems in African countries.

This seminar topic researches current energy systems, with the objective to extract the main differences between energy systems in the Western world and Africa. In a second step, prevailing energy system models shall be analyzed regarding the question in how far they fulfil the requirements in African countries, and which challenges are met by energy system models for Africa.

9) Smart Microgrids and Predictive Scheduling

Supervisor: Sokol Makolli

With the trend towards microgrids and distributed renewable energy sources, smart energy systems must manage and optimize the locally generated energy. Since renewable energy sources are inherently inflexible compared to traditional energy sources, demand response is a hot topic in energy load control for microgrids. The idea is to shift energy demand to periods of abundant energy supply. This raises a scheduling problem that requires, first, to predict energy supply and, second, to determine an optimal schedule for turning appliances off and on.

The objective of this seminar topic is to introduce a smart grid architecture and present common algorithms for energy supply prediction (e.g., AI training models, regression models) and energy demand scheduling (e.g., linear programming, stochastic programming, meta heuristics).

Melanie Heck