Network monitoring optimization : problem, models and solutions

Abstract

As data traffic grows, service providers are faced with the challenge of how to best use their infrastructure while trying to obtain statistics about their environment. The existing solutions to monitor such networks are no longer effective in these complex environments. An alternative that has been recently emerging as a promising option, allied with programmable networks, is In-Band Network Telemetry (INT), which allows for timely, accurate and fine-grained information retrieval regarding network device state. Although it has the potential to properly handle modern networks, when not employed optimally, INT can lead to high overheads and significant performance degradation. State-of-theart proposals have focused on devising new monitoring constructs and mechanisms, but have not looked at the problem from an optimization point of view. In this work, we formalize this challenge as an optimization problem, called Network Monitoring Optimization (NEMO) problem, and propose two mathematical models to solve it. The first model is based on a custom approach which aims to cover all programmable device interfaces in topologies while causing minimal overheads. The second model is presented as a generalization of the Vehicle Routing Problem. The results obtained in the model evaluation show that the proposed models can generate solutions that provide accurate and fine-grained information regarding network state while achieving minimal overheads.