RouteNet: Leveraging Graph Neural Networks for Network Modeling and Optimizat...

杨利霞

RouteNet: Leveraging Graph Neural Networks for

! E0 A1 u( m0 D- ?  L; U

Network Modeling and Optimization in SDN

5 l4 G1 K! s% F8 X& a

, }) n& b0 `) O+ _4 K, jNetwork modeling is a key enabler to achieve
% V% ~+ T: e% }2 b8 \effificient network operation in future self-driving Software
8 s- p  ]: R6 E$ U& H0 MDefifined Networks. However, we still lack functional network
models able to produce accurate predictions of Key Performance
, G5 ^1 b& R9 t& N% o# P, F. PIndicators (KPI) such as delay, jitter or loss at limited cost.
- W& a# }/ q1 f7 E  S( M( t! Y' @In this paper we propose RouteNet, a novel network model based
on Graph Neural Network (GNN) that is able to understand
; N9 S# S, Q# Gthe complex relationship between topology, routing, and input
. t4 Z2 A8 t% q2 N0 Ytraffific to produce accurate estimates of the per-source/destination
per-packet delay distribution and loss. RouteNet leverages the
ability of GNNs to learn and model graph-structured information
; U  |) H) k, A4 x. gand as a result, our model is able to generalize over arbitrary
topologies, routing schemes and traffific intensity. In our eval
7 N% `! X3 C* [" r$ q% X% ]uation, we show that RouteNet is able to predict accurately
the delay distribution (mean delay and jitter) and loss even in
topologies, routing and traffific unseen in the training (worst case
MRE = 15.4%). Also, we present several use cases where we
/ A$ _% k5 g4 p7 O" kleverage the KPI predictions of our GNN model to achieve
8 i& h" u- m0 ~0 i) U/ Ceffificient routing optimization and network planning.

0 o' Q$ l& Z& B) T1 b9 U