Interview with Software Campus Alumni Stefan Schneider
What was the content of your IT project and who were your partners?
My Software Campus IT project, RealVNF (https://realvnf.github.io/), ran for a good two years from the end of 2018 to the beginning of 2021 with Huawei Germany as a partner.
The project was about the improved coordination of chained network functions (Virtual Network Functions = VNFs) under realistic conditions. Services for video or music streaming, for mobile communications, and communication (e.g. 5G, 6G) or even for machine learning, for example, often consist of several such functions. The individual functions are software that can run on different data centers with different amounts of resources as required. The coordination of these services from several chained functions is about scaling the individual functions dynamically and continuously in an optimized manner, distributing them across different data centers and dividing up the user requests appropriately so that the service quality is as high as possible (e.g. through short latency) but no resources are wasted. The challenge is to react quickly to changing requests and to weigh up the various, sometimes conflicting optimization goals against each other. Details of the functions and requests are often unknown.
In my project, RealVNF, a deep reinforcement learning approach (a type of machine learning) was developed. The approach coordinates services in a self-developed simulation environment and learns from its own experience, without detailed instructions or expert knowledge. It adapts independently to different services, request patterns, network structures and situations and reacts quickly to changes. By learning and adapting independently to different situations, the approach is much easier to apply in practice than previous approaches and also achieves even better results (higher service quality).
The regular exchange with Huawei helped to better understand the problem and the practical challenges and to develop a corresponding flexible approach.
What was the biggest challenge you have had to face in your IT career so far?
Developing the self-learning approach with deep reinforcement learning in my Software Campus project, Real VNF, was certainly one of the biggest challenges so far – both technically and from a project management perspective.
Technically, the approach was one of the first self-learning ones in the field and it took a long time (almost 1 year) until we had designed and implemented it correctly so that it could reliably coordinate services in different situations and networks. We had to develop both the approach itself and the simulation from scratch, continuously coordinating the requirements with our partner Huawei. Building and leading the team for the project was a particular challenge: How do you get the right people? How do you divide the tasks correctly? How do you find a solution and keep the team motivated when all previous attempts have failed?
In the end, I not only learned an incredible amount from the project, but the approach developed also worked very well. We even won a Best Student Paper Award for the publication.
What is your heart beating for – besides your job?
Climate protection is very important to me. I hope that with my work at Digitale Schiene Deutschland (part of Deutsche Bahn) I can contribute to the transport transition and make railway traffic more punctual and attractive.
In my free time, I really enjoy traveling and hiking. I enjoy getting to know new countries and cultures. When I hike, I enjoy the peace and beauty of nature – and often get good, new ideas that also help me at work.
Source language of this interview: English and German
