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

- H+ h9 }! B6 k% y1 ]7 w6 {6 [
Network Modeling and Optimization in SDN
2 e) s$ I8 o- ^$ I
. l! N6 s* ^: I4 K$ H, O2 b

) n& [; b; [7 g  MNetwork modeling is a key enabler to achieve. k/ ~7 q0 P3 ]4 \
effificient network operation in future self-driving Software# c- r" z1 A9 `& f" i3 t# b
Defifined Networks. However, we still lack functional network
" F6 e0 d; l9 @0 f8 Smodels able to produce accurate predictions of Key Performance# }+ C/ R7 t0 J; T8 @
Indicators (KPI) such as delay, jitter or loss at limited cost.
3 j# }6 q, w4 v7 ?* O1 e8 uIn this paper we propose RouteNet, a novel network model based% r$ ]" x/ T+ }+ a) R- ~/ h
on Graph Neural Network (GNN) that is able to understand! |' |) f! e( J  O1 e. l& u
the complex relationship between topology, routing, and input/ p; N! U) }5 z+ D+ ^3 k( N& \- q
traffific to produce accurate estimates of the per-source/destination
: X( W5 w* A6 L5 h0 Fper-packet delay distribution and loss. RouteNet leverages the3 N% S% Z! }8 `& d8 b- b
ability of GNNs to learn and model graph-structured information; v: ~* P: m7 g5 j- @" R- `
and as a result, our model is able to generalize over arbitrary
# l( L& e5 e' g# mtopologies, routing schemes and traffific intensity. In our eval
- ~/ a+ _) m5 I2 yuation, we show that RouteNet is able to predict accurately- N" @# X, S6 K2 A6 m, o
the delay distribution (mean delay and jitter) and loss even in' ]; X+ g% S6 n" U, r- s
topologies, routing and traffific unseen in the training (worst case
6 R8 V- z9 n6 l4 ~! a/ QMRE = 15.4%). Also, we present several use cases where we5 r5 |8 ~3 L' U+ u1 p) g% C- N2 j
leverage the KPI predictions of our GNN model to achieve# P$ P  H) x- Q( F# X. X- V
effificient routing optimization and network planning.
. [5 c5 h2 Z9 W
1 \+ d) @; ]) e1 M! e) L
) ~  a; }4 d$ N% c

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