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

杨利霞

RouteNet: Leveraging Graph Neural Networks for

+ g9 T' ^1 E' r) ~) f4 R. X% z: O5 k

Network Modeling and Optimization in SDN

* G  r* C8 D) P2 i* I9 c, F* t
! T7 X+ e: `6 j
. j. _9 ~" P/ D/ W5 uNetwork modeling is a key enabler to achieve
effificient network operation in future self-driving Software
Defifined Networks. However, we still lack functional network
* H4 F" p. e/ v7 Z! rmodels able to produce accurate predictions of Key Performance
) s: T+ f! Q" g2 [4 jIndicators (KPI) such as delay, jitter or loss at limited cost.
7 Y; X8 M; f6 T1 w# K$ i, PIn this paper we propose RouteNet, a novel network model based
5 A! R) D: x4 G/ ^6 M4 u9 Uon Graph Neural Network (GNN) that is able to understand
$ }3 s& f3 |9 ]3 y; _the complex relationship between topology, routing, and input
% H7 l" q' ]  ptraffific 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
and 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
" r+ {9 v- S' k% dthe 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
leverage the KPI predictions of our GNN model to achieve
2 s9 A% z7 w( p/ zeffificient routing optimization and network planning.