A Measurement Study of a Single-BSS Software Defined WiFi Testbed
Tausif Zahid, Fouad Yousuf Dar, Xiaojun Hei, Wenqing Cheng
School of Electronic Information and Communications
Huazhong University of Science and Technology, Wuhan, China, 430074
e-mail: tausif@hust.edu.cn, fouad@hust.edu.cn, heixj@hust.edu.cn, chengwq@hust.edu.cn
Abstract—The emerging software defined networking (SDN)
provides a new paradigm of network management by
separating control functionalities from underlying data
forwarding hardware to an external controller. SDN provides
high flexibility to control the whole network in a unified way. In
this paper, we are motivated to study the performance issues of
a software defined WiFi network testbed. This testbed consists
of a single basic service set (BSS) managed by an SDN
controller. We conducted a measurement study of this testbed
under different scenarios in the lab environments. This
evaluation study focused on the throughput performance
between different devices. The testbed was evaluated under
different traffic loads in terms of nodes and packet sizes for
both TCP and UDP traffic flows. For the comparison purpose,
the performance of the conventional WiFi network is set as the
benchmark. Our results demonstrate the trade-off between
performance and flexibility of software defined access points.
Keywords-software defined WiFi networks; performance
evaluation; measurement
I. INTRODUCTION
Wireless local area networks (WLANs), also known as
WiFi networks, have become one of the key access network
technologies in the past decade. The demands on WLANs for
functionalities and scalability are growing due to the rapid
proliferation of mobile devices and applications. The number
of devices and connections per user is steadily increasing.
The dramatic increase in use of WiFi networks suggests to
improve the WiFi network infrastructure. The traditional
WiFi infrastructure is not suitable for coping with increasing
demands. This conventional mode of operations has increased
complexity, and the operational costs of running a network
have slowed down the innovation process [1].
Today’s traditional networks mostly ignore some specific
needs of users and applications. Radio resource abstraction
and allocation is not straightforward in wireless network. It is
due to the inherent broadcast nature of wireless
communications and stochastic fluctuation of wireless
channel quality. In addition, a large number of intelligent
devices with self-adaptation awareness capabilities induce
non-trivial security challenges to wireless networks. These
challenges need to be tackled broadly by comprehensive
research efforts. The networks must be programmable enough
to transmit real-time data efficiently and reliably, to aggregate
resources scalably and flexibly, to provide management
interfaces for upper layer application systems securely and
conveniently [2].
Traditional wireless networks face many issues in term of
users and network performance. User densities are not evenly
distributed over the entire space and the RF dynamics of the
AP are very different from those experienced at the user level.
The APs are exposed with an excellent view of the room and
the user devices will be packed closely together with
attenuating bodies surrounding them. The biggest factor that
affects the client devices more than any other factors is the
degradation of signal-to-noise ratio (SNR) through both co-
channel and adjacent channel interference driven by co-
located devices, user activity, percentage of users active at a
given time and the type of services they require. To address
the challenges of WiFi networks, we first need to gather the
statistics of access points of wireless local area network.
These efforts to evaluate the overall performance of a WLAN
will be greatly helpful in the optimization of a wireless local
area network. This research aims to evaluate the performance
of a software defined Wi-Fi network in terms of network
throughput and maximum achievable throughput to determine
the performance of SDN based WiFi network.
II. BACKGROUND
Software Defined Networking is a new paradigm of net-
working which promises easier management of network with
cheap hardware. Over the years, the term Software Defined
Networking has been broadened and has not become limited
to wired network. Now the SDN ideas have also been
implemented in wireless and cellular networks [3]. SDN is
also being used to optimize back-haul networks. This goal is
achieved by separating the control plane from data plane
elements and centralizing it. All the forwarding decisions are
made in a centralized location called the controller. In this
way we have a full view of the network and we can
manipulate the flow tables of these forwarding elements from
the centralized controller. An important feature of SDN-
enabled local area network is network virtualization. This is
the ability to slice the network, based on users, subnets, or
flows of traffic which allows many benefits. The isolation
provided by slicing enables one to run experiments on the
production network safely without affecting production traffic
[4]. Network virtualization has been extended to mobile and
wireless network scenario. WNV enables several concurrent
virtual networks running on the shared wireless physical
substrate. In WNV, the physical nodes and physical links are
virtualized into several virtual nodes. Correspondingly, it is
not easy for WNV to dynamically schedule the resources,
configurations, and protocols. The prospects of implementing
144
2016 First IEEE International Conference on Computer Communication and the Interne
978-1-4673-8515-2/16/$31.00 ©2016 IEEE