Research Project

Software-defined Networking



Software-defined Networking (SDN) is a current trend in communication networks based on the concepts of control plane and data (forwarding) plane separation, and logically centralized control. A logically centralized controller configures the forwarding tables (also called flow tables) of switches, which are responsible for forwarding the packets of communication flows. SDN promises several advantages such as flexibility without sacrificing forwarding performance, high efficiency through optimized routing, ease of implementation and administration, and cost reduction. With the OpenFlow standard, a protocol with strong support from industry is already available today to implement SDN. Due to these advantages, we believe that SDN will have strong influence on networking infrastructures and protocols in the near future.


Our current research in the field of SDN is focused on the following basic research questions:

How can communication middleware systems such as publish/subscribe systems and event services benefit from SDN? Today, these systems are typically implemented on the application layer in an overlay network on top of the IP forwarding infrastructure due to the basic assumption that it is not practical to modify the forwarding tables of switches or routers operating on the network layer. We argue that with the advent of SDN, this basic assumption has changed, and with suitable concepts we can implement more efficient systems benefiting from network-layer forwarding support offering line-rate throughput and low-latency messaging. [1]

Control plane distribution: SDN is based on the notion of a logically centralized controller. However, due to availability and scalability requirements, this logically centralized controller has to be physically distributed to several machines. This raises several questions that can be summarized as the "CAP problem of SDN": How can we implement a distributed controller that ensures consistency (C), availability (A), and partitioning tolerance (P)? It is well-known from distributed systems that it is hard to achieve all three CAP properties at the same time. Our goal is to find the right spot in the CAP triangle that allows for a practical solution for SDN. 

Moreover, we utilize SDN technologies to implement further distributed systems. For instance, we proposed a concept for implementing elastic and energy-efficient machines for cloud computing infrastructures by transparently integrating System-on-a-Chip (SoC) machines and classic virtual machines through SDN technologies [2].

Further details can be found in the papers listed below.


[1] Boris Koldehofe, Frank Dürr, Muhammad Adnan Tariq and Kurt Rothermel. The Power of Software-defined Networking: Line-rate Content-based Routing Using OpenFlow. In Proceedings of the 7th MW4NG Workshop of the 13th International Middleware Conference 2012.

[2] Frank Dürr: Improving the Efficiency of Cloud Infrastructures with Elastic Tandem Machines. In: Proceedings of the 6th IEEE International Conference on Cloud Computing (Cloud 2013), Santa Clara, CA, USA, June 2013

[3] Muhammad Adnan Tariq, Boris Koldehofe, Sukanya Bhowmik and Kurt Rothermel. PLEROMA: A SDN-based High Performance Publish/Subscribe Middleware. To appear in Proceedings of the ACM/IFIP/USENIX Middleware Conference.

[4] Frank Dürr: Towards Cloud-assisted Software-defined Networking. Technical Report 2012/04, Institute of Parallel and Distributed Systems, Universität Stuttgart, 2012.

This image shows Frank Dürr

Frank Dürr

Dr. rer. nat.

Senior Researcher and Lecturer

This image shows Kurt Rothermel

Kurt Rothermel

Prof. Dr. rer. nat.

Former Director of IPVS

To the top of the page