Performance analysis of software-based network functions running in different isolation environments

OData support
Dr. Csikor Levente
Department of Telecommunications and Media Informatics

In the last decades the network functions (hardware switch, router, firewall, etc.) became the fundamental part of today's information networks. These network functions, beside their efficency, had many disadvantages (bad scaling, manufacturer and model dependent configuration, etc.). According to the recently developed SDN approaches, the working mechanism of telecommunication networks can be improved by centralizing the intelligence found in the specialized network devices, compared to the performance that can be achieved by using dedicated device to each type of task. This centralized working model coincides with cloud based computing, which is quite developing nowadays, and with the virtualization technologies serving as the base of cloud computing. By centralizing the hardware functions, the tasks made by these function can be executed by software functions. The substitution of hardware functions by software functions is called Network Function Virtualization (NFV) too.

The purpose of this master's thesis is to compare the newer technologies (software based and virtualized network functions), realized by the centralization of network intelligence, with the classical solutions. For this purpose it is necessary to produce comparison capable measurements between networking functions and devices installed in different environments and implemented by different methods, especially to produce measurements for technologies, which are not commonly used in the industry yet (e.g. virtualized environment). There are very few practical measurement results about these newer technologies. Before making the measurements in the laboratory, it is necessary to work out a measurement methodology, which is capable of comparing greatly different technologies. This methodology will classify the technologies by parameters based on transmission capacity, packet delay and power consumption.

During this research work, the primary field of the investigations will be the software based network functions (e.g. layer 2 switching functions) running in virtualized, more precisely in software container (shared kernel virtualization) environment. Contrary the classical (full) virtualization, in case of the shared kernel virtualization, the virtualized systems run just partially separated, but the virtualization itself requires much less resources. In software container environment the softwares, which are implementing the network intelligence, can run centralized, but are separated from each other, which means a kind of security. This research work will include the advantages and disadvantages of the previously mentioned technologies.


Please sign in to download the files of this thesis.