|
|
OpenFlow ResearchersLead Researcher
Student Helpers
Thomas Höhn
Students
Michael Klopf
Zsolt Magyari
Christopher Metter
|
Research Field Description
The idea of OpenFlow is to separate the data path from the control path. This means that simple forwarding still resides on the switch, whereas high-level routing decisions are done on a separate controller. This split of the switching and controlling logic enables the creation of vendor-independent intelligent control mechanisms.
Modeling of OpenFlow
Understanding the performance and limitations of the basic OpenFlow concept is a prerequisite for using it for experiments with new protocols and mechanisms. Therefore, we aim to provide a performance model of an OpenFlow system. The model is based on results from queuing theory and is verified by simulations and measurement experiments with a real OpenFlow switch and controller. We intent to gradually refine the model and adapt it to the needs of the community. The advantage of an analytical model over the simulation is the fact that it can provide indicators for performance and scalability in a few seconds. time whereas the simulation may require several hours to complete depending on the computing hardware and therefore provide fast feedback.
IPOM: Interactive Proxy Management
The Interactive PrOxy Management (IPOM) tool enables us to define and emulate networks of arbitrary complexity on top of existing experimental facilities by means of OpenFlow and network emulating proxy nodes. The IPOM tool is split into two parts, the topology editor for creating a network topology and the topology management tool for controlling the flows in the network. Before running experiments in a testbed environment, the physical network topology can be mapped using the IPOM topology editor. The IPOM management tool provides the possibility to dynamically add and remove proxies. Additionally, arbitrary OpenFlow actions can be installed for any flow.OpenFlow in G-Lab
OpenFlow is quickly becoming the most dynamic topic in networking today. Therefore, it fits very well into G-Lab and its mission to create and test ideas for the networks of the future. OpenFlow extends the G-Lab experimental facility by the ability for software defined networking (SDN) and network virtualization. This enables the OpenFlow-enabled G-Lab site in Würzburg to "break the star" using the G-Lab provisioning tool ToMaTo. This provides researchers with dynamically created topologies utilizing all available server resources accelerating the creation and deployment of network experiments.
ECDC: Energy Efficient Data Center
Another application area for OpenFlow is its usage in data centers. The ability of monitoring and managing of flows provides several possibilities within data centers. In Würzburg, we are especially interested in evaluating the trade-off between energy-efficiency and Quality of Experience (QoE) within data centers. To evaluate this trade-off, we implemented the below described approach in our testbed and are currently working on the validation and improvement of the approach via simulation.
We use the Beacon OpenFlow controller to monitor the load of the servers and the data center network as well as the power consumption of the different components. Based on these information, the management entity behind the OpenFlow controller decides whether a migration of virtual machines are useful to shutdown different servers and thus, save energy. The second option is to start new servers to balance the load and to decrease the service time. The monitoring of the loads as well as the flow redirection is handled using OpenFlow.
Publications
|
Research Interests
|
Funded Projects |
Diploma and Bachelor Theses
|
|
Internships
|
|
back to NGN Group

