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

* Y. W& [; ]; T, ?
; b' }$ `4 m8 c1 m+ g) f
2 M3 F* f5 r) v2 FNetwork modeling is a key enabler to achieve
effificient network operation in future self-driving Software
5 C" p) q- W; H5 {  R9 `9 WDefifined Networks. However, we still lack functional network
" h- m5 H1 d* t& X, x* emodels able to produce accurate predictions of Key Performance
! P" E8 y/ H: R, pIndicators (KPI) such as delay, jitter or loss at limited cost.
In this paper we propose RouteNet, a novel network model based
' w0 l% \# F' O" W6 Y8 `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
3 K7 J# P1 @. Z7 o3 T( Vper-packet delay distribution and loss. RouteNet leverages the
/ o. m" `& o7 L5 k- iability of GNNs to learn and model graph-structured information
3 S# G; _3 F  q0 l7 Vand as a result, our model is able to generalize over arbitrary
+ k' U5 {; e0 j7 l8 y% }4 l: l. Ttopologies, routing schemes and traffific intensity. In our eval
8 Y% s3 p$ ^. K- nuation, we show that RouteNet is able to predict accurately
" I* I% M" E% E$ d. w/ ?the 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
effificient routing optimization and network planning.

  j) U% ^  U  @
3 i4 l2 ~, G/ Q0 o