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

杨利霞

RouteNet: Leveraging Graph Neural Networks for

) E0 s7 A: q" |) c. H' z

Network Modeling and Optimization in SDN

3 M8 b7 P! C$ m4 q7 KNetwork modeling is a key enabler to achieve
5 {0 K! i0 L8 H( Eeffificient network operation in future self-driving Software
Defifined Networks. However, we still lack functional network
) H& p# a6 K" ~) G( F+ _models able to produce accurate predictions of Key Performance
$ R/ u6 l- P" N% T( x) sIndicators (KPI) such as delay, jitter or loss at limited cost.
! _  F; y; f! UIn this paper we propose RouteNet, a novel network model based
on Graph Neural Network (GNN) that is able to understand
4 C" w; A' V! M/ _% Cthe complex relationship between topology, routing, and input
0 ^, W$ b4 x% P- r! o; u! ytraffific to produce accurate estimates of the per-source/destination
4 I; H1 d* g# Q: {6 v/ bper-packet delay distribution and loss. RouteNet leverages the
0 h4 k$ {4 @& ~ability of GNNs to learn and model graph-structured information
and as a result, our model is able to generalize over arbitrary
/ {1 O- u9 Y) d) p9 c& ?2 Q; w0 `topologies, routing schemes and traffific intensity. In our eval
uation, we show that RouteNet is able to predict accurately
# z. E' O  y+ D& H" Vthe delay distribution (mean delay and jitter) and loss even in
topologies, routing and traffific unseen in the training (worst case
- _  r( s& z6 H" xMRE = 15.4%). Also, we present several use cases where we
; R8 ?  c1 W4 y# }leverage the KPI predictions of our GNN model to achieve
effificient routing optimization and network planning.

' h6 O6 p4 @5 r* Q$ W7 F0 W) O
- a# l; K6 u* V8 b; t