Abstract

Network Functions Virtualization (NFV) is among the latest network revolutions, bringing flexibility and avoiding network ossification. While NFV provides a flexible way of implementing network functions on commodity hardware, an all-software NFV implementation may present a performance gap with respect to hardware-based solutions. In the last decade numerous software acceleration techniques have appeared to bring high-speed capabilities to software network frameworks, thus trying to reduce the distance w.r.t. pure hardware solutions. Batching is one example of such techniques, consisting in processing packets in groups as opposed to individually, which is required at high-speed to minimize the framework overhead, reduce interrupt pressure, and leverage instruction-level cache hits. Whereas several system implementations have been proposed and experimentally benchmarked, the scientific community has so far only to a limited extent attempted to model the system dynamics of modern NFV routers exploiting batching acceleration. We fill this gap by proposing a simple generic model for such batching-based mechanisms, which allows a very detailed prediction of highly relevant performance indicators. These include the distribution of the processed batch size as well as queue size, which can be used to identify loss-less operational regimes or quantify the packet loss probability in high-load scenarios. In this talk I will present our experimental campaign for performance evaluation using a state-of-the-art NFV router, namely VPP. Then I will introduce our model for a generic NFV router. We contrast the model prediction with experimental results gathered in our testbed, showing that the model not only correctly captures system performance under simple conditions, but also in more realistic scenarios in which traffic is processed by a mixture of functions. rials will be discussed.

Bio

Leonardo Linguaglossa is currently a post-doctoral researcher at Telecom PairsTech (France) working in a collaboration with Cisco named "NewNet@Paris". In 2018/2019 he is working in a joint collaboration between TPT and TUM with a project named "AI4P" (Artificial Intelligence for Performance). Leonardo's main research interests include high-speed networking, future network architectures (NFV, SDN), performance evaluation and modeling.