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

杨利霞

RouteNet: Leveraging Graph Neural Networks for

  ^* G' {( X1 I; a/ M+ S+ _; W' M

Network Modeling and Optimization in SDN

$ B* R! J5 q* B  E- X7 E* a

0 [' G5 t& |* j& z: eNetwork modeling is a key enabler to achieve
& _0 b# M& o9 }4 J' g! N0 @& M# F! beffificient network operation in future self-driving Software
  O/ ?, X$ F+ @  L- FDefifined Networks. However, we still lack functional network
models able to produce accurate predictions of Key Performance
5 f; U$ G4 o+ FIndicators (KPI) such as delay, jitter or loss at limited cost.
2 k0 A# K! i9 {In this paper we propose RouteNet, a novel network model based
) S+ l3 J9 |+ X9 ~0 @9 }  bon Graph Neural Network (GNN) that is able to understand
* G: z5 X8 Q) M3 b) Fthe complex relationship between topology, routing, and input
traffific to produce accurate estimates of the per-source/destination
per-packet delay distribution and loss. RouteNet leverages the
+ t) i7 {0 B( r% Aability of GNNs to learn and model graph-structured information
- v) M% }; }2 sand as a result, our model is able to generalize over arbitrary
topologies, routing schemes and traffific intensity. In our eval
uation, we show that RouteNet is able to predict accurately
the delay distribution (mean delay and jitter) and loss even in
4 o+ G8 P& v9 W) K$ H8 `/ Ytopologies, routing and traffific unseen in the training (worst case
MRE = 15.4%). Also, we present several use cases where we
leverage the KPI predictions of our GNN model to achieve
effificient routing optimization and network planning.

  Q7 U4 c: h, r
! t; }* @7 h2 M' f3 u; b