Exploring SDN to deploy flexible sampling-based network monitoring

Abstract

In recent years we witnessed the arrival of new trends, such as server virtualization and cloud services, an increasing number of mobile devices and online contents, leading the networking industry to deliberate about how traditional network architectures can be adapted or even deciding if a new perspective for them should be taken. SDN (Software-Defined Networking) emerged under this framing, opening a road for new developments due to the centralized logic control and view of the network, the decoupling of data and control planes, and the abstraction of the underlying network infrastructure from the applications. Although firstly oriented to packet switching, network measurements have also emerged as one promising field for SDN, as its flexibility enables programmable measurements, allowing a SDN controller to manage measurement tasks concurrently at multiple spatial and temporal scales. In this context, this paper is focused on exploring the SDN architecture and components for supporting the flexible selection and configuration of network monitoring tasks that rely on the use of traffic sampling. The aim is to take advantage of the integrated view of SDN controllers to apply and configure appropriate sampling techniques in network measurement points according to the requirements of specific measurement tasks. Through SDN, flexible and service-oriented configuration of network monitoring can be achieved, allowing also to improve the trade-off between accuracy and overhead of the monitoring process. In this way, this study, examining relevant SDN elements and solutions for deploying this monitoring paradigm, provides useful insights to enhance the programmability and efficiency of sampling-based network monitoring.