A benchmarking methodology for virtualized packet core implementations

Abstract

With the upcoming network function virtualization (NFV), aiming at deployment of core network functionality on top of cloud infrastructures, there is a stringent need for developing an appropriate evaluation process and the afferent automatic testing solutions. Before NFV, only black-box testing was needed, due to the physical isolation of the components in separate boxes. With NFV, the network components are software-only and share the same infrastructure. In NFV, a new benchmarking methodology has to be consider, accounting also for the infrastructure usage. This paper introduces such an evaluation methodology and a practical implementation, enabling the automatic evaluation how a specific virtualized packet core solution (vEPC) behaves on top of a specific data center with a specific network usage pattern. It enables the quantitative evaluations of the system from external perspective, testing how the service is perceived by the end subscribers, correlated with the system resources usage. Furthermore, the benchmarking methodology was applied and adapted based on the measurements done for the Fraunhofer FOKUS Open5GCore toolkit, a vendor independent software-only core network, through this engineering an efficient benchmarking process and giving an example of knowledge which can be extracted using such tools

A Study of 5G Edge-Central Core Network Split Options

“…The inclusion of additional network functions at the edge of the network is dependent on the available compute resources. Previous benchmarks have shown that the 3GPP systems can be considered linearly scalable for a very large number of subscribers due to the almost complete independence of the subscriber procedures [16]. This implies that with more subscribers, more compute resources will have to be allocated to the network functions at the edge and obviously, with more network functions at the edge, even more compute resources are needed.…”