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

. ]: C3 U! F/ u4 A
Network modeling is a key enabler to achieve
5 x9 h. v( T9 W% x/ s$ O' J7 \effificient network operation in future self-driving Software
" @. h! o# @% c0 a4 M) l+ qDefifined Networks. However, we still lack functional network
( U$ a& _% N% }* J) s8 wmodels able to produce accurate predictions of Key Performance
/ u( b+ x( n$ r4 H3 gIndicators (KPI) such as delay, jitter or loss at limited cost.
& Q. |% ?4 o3 w. c' N4 W1 u! L: KIn this paper we propose RouteNet, a novel network model based
8 N! L8 e6 V- ~4 ~on Graph Neural Network (GNN) that is able to understand
the 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
ability of GNNs to learn and model graph-structured information
- U8 A* j! ~1 ]! I2 J9 Jand as a result, our model is able to generalize over arbitrary
topologies, routing schemes and traffific intensity. In our eval
/ s4 J' u3 \% {  g: B! Z5 Uuation, we show that RouteNet is able to predict accurately
the delay distribution (mean delay and jitter) and loss even in
& J8 [, G" N( l# a  @2 d, Htopologies, routing and traffific unseen in the training (worst case
MRE = 15.4%). Also, we present several use cases where we
& [$ k% ?! J$ G7 Q/ @' y- oleverage the KPI predictions of our GNN model to achieve
effificient routing optimization and network planning.

/ w7 R  a" \8 R8 v
: f3 j6 U8 l0 @- w