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

% r0 I, c) |" ?* r) _: d3 ^Network modeling is a key enabler to achieve
  z' W2 z* F* {3 i9 F9 ~$ d! ieffificient network operation in future self-driving Software
% F: v$ F# ], j' p2 a. p! ZDefifined Networks. However, we still lack functional network
models able to produce accurate predictions of Key Performance
  M; B0 e# B6 b2 WIndicators (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
# A0 u# t& [9 i5 U( w. Ethe complex relationship between topology, routing, and input
traffific to produce accurate estimates of the per-source/destination
$ e7 j( i8 k! fper-packet delay distribution and loss. RouteNet leverages the
1 B9 d% c0 E1 z* x' e- u( Jability of GNNs to learn and model graph-structured information
' q3 Q- Q: \0 Y6 cand as a result, our model is able to generalize over arbitrary
topologies, routing schemes and traffific intensity. In our eval
1 K- P. m, t, x& N$ vuation, we show that RouteNet is able to predict accurately
: X% q+ n+ C. ^& mthe delay distribution (mean delay and jitter) and loss even in
topologies, routing and traffific unseen in the training (worst case
( V- G) ^1 \0 ?* e  eMRE = 15.4%). Also, we present several use cases where we
( S8 G' d( @% C; d& q/ g! X6 ileverage the KPI predictions of our GNN model to achieve
2 K3 i5 p* S  ?% r. x- c2 P/ g& ?effificient routing optimization and network planning.

2 X* M: ]" G$ z, a  Y  v# S2 v
* E- Q8 P$ ]* L2 w1 S) K