Effective Capacity Analysis of Multiuser
Ultra-Dense Networks with Cell DTx
Qing Li
∗
,YuChen
∗
,QimeiCui
∗
,YuGu
∗
and Guoqiang Mao
†
∗
National Engineering Laboratory for Mobile Network Technologies, Beijing University of Posts and Telecommunications
Email: {yu.chen, cuiqimei, liqing1994}@bupt.edu.cn
†
School of Computing and Communications, University of Technology Sydney
Email: g.mao@ieee.org
Abstract—Cell discontinuous transmission (Cell DTx) im-
proves network performance because it help mitigate inter-cell
interference. However, the relationship between this technology
and network performance has been little studied. The aim of
this work is to understand the impact of Cell DTx on users’
quality of service (QoS) performance of multiuser ultra-dense
networks (UDNs). We extend the traditional one-dimensional
effective capacity model and develop a new multidimensional
framework for UDNs with Cell DTx, which can be applied to
different scheduling policies. Under the round-robin and the
max-C/I scheduling examples, computer simulation shows that
the analytical and simulation results are in good agreement and
thus validate the accuracy of our proposed framework.
I. INTRODUCTION
The cell densification is a key feature in 5G ultra-dense
networks (UDNs) to enhance network throughput [1]. When
base stations (BSs) are densely deployed and operating in the
same frequency spectrum [2], severe inter-cell interference
(ICI) is caused and a common technique to reduce ICI is
the use of the cell discontinuous transmission (Cell DTx)
technology. Moreover, 5G systems allow multiple users to
share a channel and are required to support quality-of-service
guarantees for different services. Therefore, it is important
to understand the stochastic behavior of QoS performance of
multiuser UDNs with Cell DTx.
A number of work has studied Cell DTx in improving the
physical-layer spectral efficiency [3], [4] and energy efficiency
performance [5], [6]. Polignano et al. [7] studied the impact of
Cell DTx on the link-layer QoS performance of Voice over IP
traffic under dynamic and semi persistent packet scheduling
strategies by simulations. In order to mathematically analyze
the QoS performance, such as queue length distribution and
delay-violation probability, some work adopts a cross-layer
effective capacity model [8]-[10]. Under full-buffer scenarios,
Weyres et al. [11] analyzed the effective capacity in mul-
tiuser networks under the round-robin (RR) and the max-C/I
scheduling policies. Cui et al. [12] calculated the maximum
sustainable data rate of a link in UDNs under QoS constraints.
Gu et al. [13] considered the burstiness of traffic in UDNs
and derived the effective capacity in single-user scenarios.
However, Cell DTx and multiuser scenarios are not jointly
considered in those work.
The aim of this work is to quantitatively analyze the effec-
tive capacity and QoS performance with Cell DTx in multiuser
UDNs. In particular, we continue Gu’s work and develop a
new analysis framework to approximate QoS performance of
UDNs with Cell DTx, which could be applied to multiuser
scenarios under the RR and the max-C/I scheduling policies.
The rest of this paper is organized as follows: Section II
describes the multiuser UDNs model. Preliminaries on the
effective capacity model is given in Section III. In Section IV,
we develop a multidimensional effective capacity framework
to analyze the QoS performance of UDNs with Cell DTx. In
Section V, we apply the proposed framework to derive the
approximation of the effective capacity of multiuser scenarios
under the RR and the max-C/I scheduling policies. In Section
VI, we build a simulation platform to validate the accuracy of
our proposed framework. Section VII summarizes our work.
II. S
YSTEM MODEL
A. System Architecture
A system model example of a multiuser UDN is shown in
Fig. 1. The network consists of several SBSs, each serves
multiple user equipments (UEs) in a shared channel. The
interference experienced by a UE is assumed to be caused
by the most adjacent SBS [11] , so the network can be
simplified to a two-SBS model. The set of SBSs is denoted
by 𝒩 = {1, 2}; the set of UEs in SBS 𝑛 (𝑛 ∈𝒩) is denoted
by 𝒥
𝑛
= {1, 2, ⋅⋅⋅ ,𝐽
𝑛
}, where 𝐽
𝑛
is the total number of
associated UEs.
Each UE is assigned an infinite-sized buffer in its associated
SBS. SBSs apply a specific scheduling policy to allocate the
shared time resource among multiple UEs.
All the SBSs use Cell DTx, i.e., each SBS has two modes,
namely an idle mode and an active mode [14]. If the buffer
of the scheduled UE is empty and there is no data arrival at
this slot, then the SBS switches to the idle mode; otherwise, it
stays in the active mode for the target UE’s data transmission.
Fig. 2 illustrates the transition between these modes.
B. Inter-Cell Interference Modeling
We assume that the wireless channels experience block
Rayleigh fading and channel conditions of UEs in SBS