Tailored Load-Aware Routing for Load Balance
in Multilayered Satellite Networks
Yu Wang
1
, Min Sheng
1
, King-Shan Lui
2
, Xijun Wang
1
, Runzi Liu
1
, Yan Zhang
1
and Zhou Di
1
1
State Key Laboratory of ISN, Institute of Information Science, Xidian University, Xi’an, Shaanxi, 710071, China
2
Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong
Email: yu_wang@stu.xidian.edu.cn, msheng@mail.xidian.edu.cn, kslui@eee.hku.hk
Abstract—A Multilayered Satellite Network (MLSN) tends to
be a promising architecture in facilitating global ubiquitous
broadband communication. However, unbalanced traffic distri-
bution among its satellite layers should frequently occur, where
the lower layers could get relatively congested while the upper
layers remain underutilized. This unfair distribution of network
traffic can lead to large end-to-end delay and severe throughput
degradation. To cope with the above issue, we propose a Tailored
Load-Aware Routing (TLAR) strategy to optimally distribute
traffic load among the multiple satellite layers, so that the overall
traffic congestion in the MLSN is minimized. In TLAR, an
optimal portion of network load, which is decided based upon
the newly arrived traffic estimation and theoretical analysis of
traffic congestion rate in each layer, is detoured through the
upper layer. The performance of the proposed routing method has
been validated through extensive simulations, which demonstrate
that TLAR can significantly alleviate traffic congestion, achieve
low end-to-end delay and sustain improved throughput.
I. INTRODUCTION
As a promising vision of the next generation satellite net-
works, multilayered satellite networks (MLSNs) [1] integrat-
ing multiple satellite constellations have been proposed in the
literature. The MLSN architecture exhibits enormous potential
in providing global ubiquitous broadband communication due
to various advantages, such as capacious coverage, reinforce-
ment of the network capacity, and lower delay performance
[2], [3]. Therefore, the MLSNs are envisaged to cooperate
with the terrestrial wireless networks to accommodate the
burgeoning communication demands in the near future.
In order to efficiently utilize the scarce network resources
of the MLSNs, fair traffic distribution among its multiple
layers is, indeed, crucial. However, given the highly dynamic
feature of MLSNs, along with the high variance in the user
density influenced by geographical restrictions, MLSNs should
be prepared to confront a distinct challenging scenario from
its individual layers, where the lower layers are relatively
congested while the upper layers remain underutilized. With-
out an efficient load balanced routing algorithm, this unfair
distribution of network traffic will inevitably lead to significant
cumulative queueing delay and high packet drop rate.
To cope with the issues mentioned above, several load
balancing schemes have been devised to fairly distribute traffic
This paper is supported by NSFC (91338114, 61231008, 61172079,
61201141, and 61301176), 863 project (No. 2014AA01A701), 111 Project
(B08038).
into the whole network. The authors in [4] focus on the
GEO/LEO hybrid network and deal with the load balancing
issue with the provision of quality of service (QoS). A
congestion-prediction based mechanism is employed to effec-
tively avoid the occurrence of actual network congestion. Once
congestion is predicted at a satellite, traffic is immediately
detoured to achieve load balancing before actual congestion
could happen. However, one drawback exists that the best-
effort traffic is always bypassed through the GEO layer regard-
less of the congestion situation in the whole network. In other
words, when some satellites in the LEO layer get congested
while the overall traffic in the layer is not heavy, the best-effort
traffic should be detoured via neighboring LEO satellites rather
than GEO satellites to reduce end-to-end delay.
The recent work in [2] introduces an optimal timescale
threshold for efficient traffic distribution in a two-layered
LEO/MEO network. In the proposed routing approach, long-
distance traffic whose measured end-to-end delay exceeds
the optimal timescale threshold is transferred by the MEO
layer. On the contrary, short-distance traffic is transmitted
through the LEO layer. Based on this mechanism, the LEO
layer and MEO layer are equally utilized. Nevertheless, since
the propagation delay of the MEO layer is large, for delay-
sensitive traffic, transmission over the LEO layer should be
preferred, especially when the total amount of traffic in the
MLSN is relatively low. In addition, the value of the timescale
threshold is assumed to be fixed regardless of any incident,
which weakens its adaptability to the load fluctuation.
In this paper, we concentrate on the particular problem of
load balance among multiple satellite layers in the MLSN.
A Tailored Load-Aware Routing (TLAR) scheme with low
computational complexity and light signaling overhead is
devised to minimize the occurred traffic congestion
1
.The
proposed TLAR strategy can efficiently decide on the amount
of traffic that should be distributed into each layer, according
to the periodically perceived load information. Upon collecting
such load information, we theoretically quantify the estimated
traffic congestion in the network based on the queueing theory,
and then formulate the optimal traffic distribution problem as a
convex optimization problem. Finally, the closed-form solution
1
Traffic congestion is used to characterize the degree that the total traffic
volume in the network approaches the network capacity.