数学建模社区-数学中国
标题: RouteNet: Leveraging Graph Neural Networks for Network Modeling and Optimizat... [打印本页]
作者: 杨利霞 时间: 2020-11-16 15:20
标题: RouteNet: Leveraging Graph Neural Networks for Network Modeling and Optimizat...
RouteNet: Leveraging Graph Neural Networks for
1 C$ _; Q5 e \( a p& X3 s
Network Modeling and Optimization in SDN
$ u0 f. I: m8 d6 W
0 Z& x7 B2 e1 C, } x% p/ G1 \* J/ X
Network modeling is a key enabler to achieve$ N- E& a! m8 @* ]+ j" S0 ` a
effificient network operation in future self-driving Software: g. v0 D3 x' h7 D
Defifined Networks. However, we still lack functional network; b+ n9 I* i2 R6 P) L3 n
models able to produce accurate predictions of Key Performance9 S% R5 d# x6 s- G4 ]& {
Indicators (KPI) such as delay, jitter or loss at limited cost.0 W& ]: u7 V* m0 ?# K) A
In this paper we propose RouteNet, a novel network model based1 m+ N/ ^8 P+ M4 z+ z0 R2 h% n C
on Graph Neural Network (GNN) that is able to understand4 ^, n; x! L0 Y) B
the complex relationship between topology, routing, and input
9 M+ ^; ^ ]4 }( ftraffific to produce accurate estimates of the per-source/destination
n' ^9 J6 l: X( bper-packet delay distribution and loss. RouteNet leverages the
1 j* |1 B% E' T, v7 uability of GNNs to learn and model graph-structured information
8 q y1 j" G Z% e$ U+ k) c% n$ Yand as a result, our model is able to generalize over arbitrary( z( |& ]2 I5 W1 I
topologies, routing schemes and traffific intensity. In our eval5 m, N6 i) C3 E% c
uation, we show that RouteNet is able to predict accurately$ h h3 Y4 S+ t. \9 h
the delay distribution (mean delay and jitter) and loss even in* V+ |: A/ L5 L% K. O! ~2 R
topologies, routing and traffific unseen in the training (worst case7 _( b+ ?8 S& B
MRE = 15.4%). Also, we present several use cases where we
- f- Z" d$ f2 p( T1 Q0 y% uleverage the KPI predictions of our GNN model to achieve
3 s& W' N Y$ `3 {7 b: |* ]effificient routing optimization and network planning.
& l' ?9 Y2 U% P: \+ r. o
+ z! ]; ]& H' N6 N# e) ~
! f! J* ]9 ~7 c' x4 N
-
-
08934670.pdf
999 KB, 下载次数: 0, 下载积分: 体力 -2 点
| 欢迎光临 数学建模社区-数学中国 (http://www.madio.net/) |
Powered by Discuz! X2.5 |