TOM: a self-trained Tomography solution for Overlay networks Monitoring

Abstract

Network tomography is a discipline that aims to infer the internal network characteristics from end-to-end correlated measurements performed at the network edge. This work presents a new tomography approach for link metrics inference in an SDN/NFV environment (even if it can be exported outside this field) that we called TOM (Tomography for Overlay networks Monitoring). In such an environment, we are particularly interested in supervising network slicing, a recent tool enabling to create multiple virtual networks for different applications and QoS constraints on a Telco infrastructure. The goal is to infer the underlay resources states from the measurements performed in the overlay structure. We model the inference task as a regression problem that we solve following a Neural Network approach. Since getting labeled data for the training phase can be costly, our procedure generates artificial data for the training phase. By creating a large set of random training examples, the Neural Network learns the relations between the measures done at path and link levels. This approach takes advantage of efficient Machine Learning solutions to solve a classic inference problem. Simulations with a public dataset show very promising results compared to statistical-based methods. We explored mainly additive metrics such as delays or logs of loss rates, but the approach can also be used for non-additive ones such as bandwidth.

Monitoring Network Slices with a Genetic Algorithm Approach

“…Statistical methods, such as the Fourier transform, were also proposed to estimate the link metric's distribution [15]. More recently, machine learning techniques have been introduced as a statistical solution for network tomography [16], [17], and [18]. In [16], the authors used a neural network to infer underlay resources from overlay network measurements, achieving accurate estimation.…”

“…More recently, machine learning techniques have been introduced as a statistical solution for network tomography [16], [17], and [18]. In [16], the authors used a neural network to infer underlay resources from overlay network measurements, achieving accurate estimation. A neural network was also implemented in [17] to infer slices' metrics.…”

“…However, these values are not needed once the paths are determined. Additionally, other tomography techniques can be used, some of which do not require knowledge of X [16].…”

On the use of machine learning and network tomography for network slices monitoring

“…Compared to our work, we also use supervised learning but we train a model using an automatically generated dataset. In [22], M.Rahali and al, used machine learning algorithms and more particularly neural networks. They trained them using a simulated training dataset.…”

Network Identifiability: Advances in Separating Systems and Networking Applications

“…These techniques are very versatile and prove to be useful in many application contexts, including urban traffic monitoring [7]. Although born before the advent of the latest centralized network frameworks (e.g., software-defined networks and network function virtualization), NT techniques can provide efficient monitoring also in these contexts with negligible traffic overhead [8], [9]. However, BNT also brings about many challenges, mostly related to the computational intractability of data interpretation.…”