[Back]

D. Djamiykova:
"Monitoring the Correspondence of Physical and Virtual Network Resources in OpenFlow Based Software Defined Networks";
Supervisor: S. Dustdar, Ph. Leitner; Institut für Informationssysteme, Arbeitsbereich Verteilte Systeme, 2013; final examination: 10-07-2013.

As cloud computing gains more popularity, the need for easy provisioning, monitoring and prediction of the underlying network increases. In the cloud environment these requirements can be
fulfilled by virtualizing the network resources. Even though network virtualization technologies already exist in the traditional networks, there is a need for new methods of network virtualization
due to the new challenges that the cloud network imposes. Software Defined Networks (SDNs) are a new paradigm for designing and implementing virtual networks by abstracting the control from the forwarding plane. SDNs promise to bring better scalability, flexibility and efficiency to the cloud network compared to those provided by the conventional model. The single production-ready implementation of the SDN idea is the OpenFlow protocol. It adopts
a centralized management model of the virtual network by isolating the functions of physical devices from those of the virtual ones. Nowadays, one of the most challenging problems of cloud network virtualisation, built upon a SDN, is the poor management and monitoring of the correspondence between the physical and the logical devices.
This work introduces an extension to an existing OpenFlow controller that aims to monitor the physical network resources and to map gathered metrics to the corresponding virtual network.
Using low-level monitoring in every cloud setup enables optimal utilization of the physical devices and better and faster provisioning of the logical network as well as of the tenant´s
VMs. The proposed OpenFlow controller plug-in enables such low-level monitoring, centralized for the network. The plug-in provides information for changes and faults in the physical network providing feedback for any enforced management policies, and helping for an effective enforcement of network QoS and eventually, for a prediction of SLA violations. The functionality that is proposed by the implemented plug-in enables comprehensive testing and optimization
of the virtual network infrastructure with regards to the physical resources.
Furthermore, an analysis of the interconnection between physical and virtual network resources is provided based on measurements gathered through continuous monitoring of the network behaviour. The methodology of the analysis includes simulations of network topologies
that implement different relations between the physical and the virtual resources. The simulation results are duplicated and verified by a test-bed with the same topology implementations. The evaluation of the different networks shows that there is no universal solution that would fit different types of network requirements.