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

杨利霞

RouteNet: Leveraging Graph Neural Networks for

& J" L7 V4 g$ `

Network Modeling and Optimization in SDN

$ @5 l! h9 P# l" W$ D" I+ [8 S- e

Network modeling is a key enabler to achieve
effificient network operation in future self-driving Software
8 U; |1 c0 Z; V4 fDefifined Networks. However, we still lack functional network
models able to produce accurate predictions of Key Performance
! G6 u: ]5 g7 f9 N0 {1 ^, H; zIndicators (KPI) such as delay, jitter or loss at limited cost.
In this paper we propose RouteNet, a novel network model based
on Graph Neural Network (GNN) that is able to understand
4 ?8 I6 D. V; A/ x. ?3 i/ Ethe 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
and as a result, our model is able to generalize over arbitrary
: n, ]8 p. l3 H; C- Ttopologies, routing schemes and traffific intensity. In our eval
4 x; a7 c. C$ u$ I& I+ b  huation, we show that RouteNet is able to predict accurately
7 R: t# J; E0 j: h! lthe delay distribution (mean delay and jitter) and loss even in
% w+ x. E& |! G! r1 Gtopologies, 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
; m6 O+ W" ?; i' J8 u. x( Jeffificient routing optimization and network planning.
1 N8 T  x/ W$ A; o