Problem: BGP is performance-agnostic

Solution: incorporate a delay-related metric into the best-path selection process.

Approach: Estimate the round-trip prop_delay to destinations (/24s) within the routing table of the IXP

Goal: Evaluate if it is possible to outperform BGP’s route selection criterion, in terms of latency, with a measurement-based approach.

Analyzing traffic in today's communication networks becomes more and more complex due to the increasing heterogeneity of networked devices and general rising traffic volumes. 
In this context, data-driven methods can facilitate a deep understanding of the inherent dynamics needed to operate such networks efficiently.

Because network-related data is often naturally represented in graph form, this Internship specifically seeks to evaluate how methods from the Pytorch library for temporal graph learning, "PyTorch Geometric Temporal," can be applied to IP-to-IP level data.  

Goal is to among others to predict:

• IP-node activity over time 
• IP-to-IP level communications (e.g. number of packets) over time

Steps:

1. Set up a working ML pipeline.
2. Fine-tune models / try different models.
3. Identify promising directions and limitations
4. Conclude what needs to be done/tried in the future.

Prerequisites