The Taub Faculty of Computer Science Events and Talks

Network function cloudification aims at executing virtual network functions on the cloud to enable rapid deployment of new services with greater flexibility and better CAPEX and OPEX efficiencies. While the cloud provides a convenient overlay abstraction to the underlying compute and network infrastructure, the data will often travel over the network to be processed by these virtual functions and possibly travel again to be consumed elsewhere. Network overlays suppress important topological details of the underlying network infrastructure essential for this routing process. This, in turn, might lead to significant congestion and performance degradations.

In this work, we take a data analytics approach and analyze end-to-end overlay networking measurements to extract knowledge and insights about the opaque underlay topology. With this knowledge, we can then optimize the performance of the underlay network by dynamically regulating the traffic transmitted over that network. This unique interdisciplinary approach combining big data analysis with classical networking paradigms proves to be very effective.

We consider two use-cases, namely network congestion control and network availability of multiple data centers. Our results demonstrate significant performance improvement for this approach compared to current traditional approaches that completely ignores the properties of the underlay network. We demonstrate these improvements in a variety of small scale and in real scale settings.