Extending a SAIN Architecture Agent with Active Network Measurements

Abstract

In a context where Intent-based Networking is in fast evolution, techniques to provide service assurance are becoming more and more important. A project named SAIN, for Service Assurance in Intent-based Networking, was started and is still in the process of standardization. The University of Liège has developed the prototype of an agent of this architecture whose objective is to determine symptoms and health levels of the different sub-services of a network service.

This agent, named DxAgent, was fully functional but did not yet have any metrics to assess the health of network connections. Therefore, the objective of this thesis is to elaborate the implementation of active measures and thus to be able to assign a health score to the network but also to determine potential symptoms.

The measurement tool that has been chosen is the One-way Active Measurement Protocol (OWAMP), which has the advantage of providing one-way measurements. A python interface of the standard implementation of this protocol was then developed to allow the retrieval of the metrics.

Then this one was integrated to the DxAgent to extend it to new metrics and thus widen the agent's field of view. This extension proved to be a success and the DxAgent gets even closer to its original goal by allowing to discover symptoms proper to network metrics (delay, reordering, link failure, etc).

Finally, a use case has been developed to show how these additions work and how useful they can be. A topology based on docker containers simulating the network of a company offering a streaming service has been set up. Moreover, a tool allowing to automate parameterized scenarios on this architecture has been developed. The results of the experiments show that the metrics and the detection of various symptoms work well in a near real-life context.