Localization of network service performance degradation in multi-tenant networks

Modern network services are in most cases virtualized and the traffic of various users is multiplexed over the same physical links. Traditional network monitoring methods which predominantly rely on the physical interface monitoring are not sufficient because they do not provide the insight into the behaviour of the traffic in the specific network service instance or per-user service experience.

This paper presents NetMon, a framework that gathers performance indicators inside the network service and at the same time allows an efficient spatial performance degradation localization capability. NetMon is technology agnostic and suitable for various network technologies ranging from legacy to the new type of services based on network function virtualization, chaining service functions or programmable network elements.

The paper presents an original solution to the problem of scalable active network service monitoring with the capability to distinguish the measurements of different virtual networks. The proposed hybrid monitoring method which mixes the active probing and capturing this probe traffic provides a good trade-off between the granularity of the results and the traffic overhead created by the monitoring system itself.

The system was tested in the pan-European GEANT network which connects European research and education infrastructures and showed reliable performance over long periods of time.