Learning to Configure Computer Networks with Neural Algorithmic Reasoning

Luca Beurer-Kellner · Martin Vechev · Laurent Vanbever · Petar Veličković

Hall J #505

Keywords: [ neural algorithmic reasoning ] [ graph neural networks ] [ configuration synthesis ] [ computer networks ] [ Systems ]

[ Abstract ]
[ Paper [ Poster [ OpenReview
Thu 1 Dec 2 p.m. PST — 4 p.m. PST


We present a new method for scaling automatic configuration of computer networks. The key idea is to relax the computationally hard search problem of finding a configuration that satisfies a given specification into an approximate objective amenable to learning-based techniques. Based on this idea, we train a neural algorithmic model which learns to generate configurations likely to (fully or partially) satisfy a given specification under existing routing protocols. By relaxing the rigid satisfaction guarantees, our approach (i) enables greater flexibility: it is protocol-agnostic, enables cross-protocol reasoning, and does not depend on hardcoded rules; and (ii) finds configurations for much larger computer networks than previously possible. Our learned synthesizer is up to 490x faster than state-of-the-art SMT-based methods, while producing configurations which on average satisfy more than 93% of the provided requirements.

Chat is not available.