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

杨利霞

RouteNet: Leveraging Graph Neural Networks for

Network Modeling and Optimization in SDN

5 V. v6 T. @4 ~1 F* d0 H) ~5 ?( a
! M* Y- ]  c  ~  L  R# A; p) }! j
Network modeling is a key enabler to achieve
! ^# Q8 B5 |% H0 _' g/ f6 Neffificient network operation in future self-driving Software
1 n7 d8 e' \) K! ^3 j( qDefifined Networks. However, we still lack functional network
6 p' J' u. q- I. Z: B9 Z5 ?models able to produce accurate predictions of Key Performance
/ l" {; e8 \& \Indicators (KPI) such as delay, jitter or loss at limited cost.
7 m5 R/ K, b( O7 ~7 m( I0 gIn this paper we propose RouteNet, a novel network model based
. ~+ d8 b7 `- L, e( e/ J( Fon Graph Neural Network (GNN) that is able to understand
" X- E' i/ d2 }! A  I7 \; lthe complex relationship between topology, routing, and input
traffific to produce accurate estimates of the per-source/destination
# j4 o& s' C: T2 u( O: W/ xper-packet delay distribution and loss. RouteNet leverages the
ability of GNNs to learn and model graph-structured information
7 B/ f8 y/ H1 w8 jand 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
) K. s5 Z; d! L" a9 tthe delay distribution (mean delay and jitter) and loss even in
1 [' c) e6 |& Dtopologies, routing and traffific unseen in the training (worst case
3 L3 C7 A' Q- FMRE = 15.4%). Also, we present several use cases where we
4 v1 `4 @/ i  ?6 A# Uleverage the KPI predictions of our GNN model to achieve
effificient routing optimization and network planning.