Вы находитесь на странице: 1из 6

Seamless Handoff Management in IEEE 802.

11
Networks Using SDN
Steven Martinez∗ , Nelson Cardona† , Juan F. Botero‡
Universidad de Antioquia, Medellı́n, Colombia
Email: ∗ steven.martinez@udea.edu.co, † nelson.cardona@udea.edu.co, ‡ juanf.botero@udea.edu.co

Abstract—The implementation of IEEE 802.11 wireless net- Network device of packet Data plane
works has grown notably in recent years. Handoff management switching link
in wireless networks, i.e., transferring an active data session from Network device with
one cell defined by an access point to another, is a critical func- Control plane
embedded control link
tion. In IEEE 802.11 networks, this function interrupts mobile
users’ connection to the network, ultimately resulting in service SDN
disruption. Currently, the Software Defined Networking (SDN) Controller
paradigm presents innovations and abstractions in the network
architecture that can improve network management in wireless
networks. SDN, through Light Virtual Access Point abstraction,
enables the network to take control of the handoff process in
IEEE 802.11 without network disconnection (seamless handoff)
which is not possible in IEEE 802.11 traditional networks. In
this paper, we show through experimentation over real network
devices that SDN enables to perform the seamless handoff process Traditional Network SDN Network
in IEEE 802.11 without modification to the end user’s mobile
(Distributed Control) (Centralized Control)
station and completely controlled by the network. We show that
it is possible to reduce the handoff delay from one second or
more in traditional networks to milliseconds in SDN networks Fig. 1. Comparison between traditional and SDN networks
by means of the LVAP abstraction.
Index Terms—IEEE 802.11, handoff management, Software
Defined Networks, SDN.
In this paper, we show that SDN allows to perform the
handoff process in IEEE 802.11 without network disconnec-
I. I NTRODUCTION tion, i.e., seamless handoff. The above is accomplished without
any modifications to the MU and completely controlled by the
Many Wireless Local Area Networks (e.g. universities, insti-
network. Through experimentation over real network devices,
tutional networks or companies) are characterized by small and
we show that the handoff delay in SDN networks is in the
crowded geographic spaces, high density of mobile users and
milliseconds scale versus one second or more in IEEE 802.11
high congested access points [1]. In general, the IEEE 802.11
traditional networks. The rest of the paper is organized as
standard is the main used technology for these networks.
follows. Section II shows the background and theoretical
In this standard, when Mobile Users (MU) move from an
aspects of this paper. Section III describes the software and
Access Point (AP) to another, they lose the connection with
hardware considerations to implement the LVAP abstraction
the network and thus with the services. This is due to the fact
in real network devices. The performance evaluation of our
that IEEE 802.11 handoff is a hard handoff, i.e, MU has to
experiments is presented in Section IV. Finally, conclusions
finish its communication session with the current AP before
and future work are presented in Section V.
being migrated to another AP. This condition implies that MU
will suffer from service intermittences and bad user experience II. BACKGROUND
[2]. In this section we present the relevant concepts that will be
Currently, Software Define Networking (SDN) presents in- used for the development of the experiments involved in the
novations in the network architecture that can improve the per- realization of this work.
formance in wireless networks. SDN separates the data plane
from the control plane, simplifying the control network and A. Software Defined Networks - SDN
enhancing the network management towards to better control SDN is an emerging architecture for telecommunications
over network [3]. Specifically in IEEE 802.11 networks, SDN networks that breaks with the current vertical integration
enables the network to take control of the handoff process scheme separating the data plane (in charge of switch-
through the Light Virtual Access Point (LVAP) abstraction and ing/routing the packets) from the control plane (in charge of
makes possible seamless handoff which is not possible in IEEE defining how it is handled the traffic) in order to facilitate the
802.11 traditional networks [4]. network management. Fig. 1 illustrates how, in SDN, network

978-1-5386-6820-7/18/$31.00 ©2018 IEEE


WTP WTP2
LVAP1
LVAP1 LVAPn LVAP2
MU1
MAC MU MAC MU SDN
IP MU IP MU Controller

BSSID BSSID
SSID SSID WTP1
MU2

Fig. 2. Stored information in a LVAP


(a)

The Controller migrates


devices are converted into switching devices (data plane) and LVAP1 from WTP1 to WTP2
WTP2
the control of the network (control plane) is performed through MU1
a logically centralized device called controller. This in contrast LVAP1
LVAP1

to traditional networks, where the data plane and the control LVAP2
plane are embedded in the same device [3]. SDN
For the communication between the devices of the data Controller
plane and control plane, a communication protocol called
OpenFlow is used [5].
WTP1

B. Handoff management in IEEE 802.11: traditional networks MU2

Handoff management is the transfer of a Mobile User (MU)


(b)
data session information from one AP to another through a
sequence of messages. It involves three phases: discovery, Fig. 3. Migration process of a LVAP
(re)-association and (re)-authentication [6]. In IEEE 802.11
networks, handoff is a hard handoff. Hard handoffs imply
service disruption between the MU and the network.
The discovery phase begins when the Received Intensity LVAP allows to simplify association, authentication and
Signal Indicator - RSSI perceived by the MU degrades un- handoff process in IEEE 802.11 networks. Furthermore, it
til it loses connection. The MU scans all APs nearby and helps to unify the wired and wireless networks. Different
sends probe requests frames in all operation channels. The MUs connected to the same AP are logically managed in
avaliable APs answer with probe responses frames. In the an isolated way. Handoff management is facilitated with the
association phase, the MU sends an association request frame LVAP because the network infrastructure is allowed to control
to the selected AP (the one that better RSSI reported) that the handoff process without making any modification in the
rensponds with an association response frame informing that MU and without triggering any link layer reassociation frame.
it is available. Once the association phase is finished, the MUs connect to the SDN network as if they were connected
authentication phase begins. Here, the AP and MU exchange to a traditional network.
the authentication request/responses frames in order to allow The LVAPs are created by the SDN controller and allocated
access to the network. in the Wireless Termination Points - WTPs. The process
Several methods have been proposed to perform handoff performed by the controller to assign a LVAP to an AP is
in traditional IEEE 802.11 networks. The main approaches summarized as follows:
to perform handoff in traditional networks are summarized First, MUs connect to the network following the standard
in [7]. Since the main goal of this paper is to compare the (traditional) handoff process. A LVAP is created for each
handoff process in IEEE 802.11 traditional networks against MU. This enables the network to take control for association
IEEE 802.11 SDN networks, next we present the main method decisions and MUs management regarding them as isolate
to perform handoff in SDN networks. clients in the IEEE 802.11 MAC layer. Fig. 2 shows the stored
information in a LVAP. For each associated MU, MAC and IP
C. Handoff management in IEEE 802.11: SDN networks addresses are stored, as well as, a unique BSSID to create the
The most important work in handoff management through illusion that each MU has its own AP and one or more SSIDs.
SDN is Odin [4], [8]. The authors propose the Light Virtual Fig. 3 shows the process to migrate a LVAP from a physic
Access Point - LVAP, an abstraction that allows to perform AP to another. In (a) MU1 and MU2 are connected to WTP1
the handoff process without service disruption, i.e., seamless through LVAP1 and LVAP2 respectively. In (b) MU1 moves
handoff. causing a degradation of the RSSI and forcing MU1 to
UDP Server The requirements of WTPs are those necessary for the
implementation of the LVAP abstraction. In this case it is
DHCP SDN necessary to comply with the requirements regarding the
Server Controller generation of ACKs and agent implementation. While the tools
used by EmPOWER are open source, Ssme of the following
functionalities must be supported by the WTPs: 1) OpenWRT
[14] operating system support, 2) frame injection support via
radiotaps [15], 3) the possibility to create a BSSID mask in
Switch
order to create a unique BSSID for each LVAP [16] and 4)
flash memory greater than 8MB.
Considering these restrictions, we proceeded to search the
WTP1 WTP2 devices that fulfill these requirements and that are easily ac-
cessible in Colombia with very good hardware specifications.
After performing an exhaustive technological surveillance,
MU we selected the APs TP-LINK Archer C7 AC1750 Version
2.0 which fulfilled all the established requirements. The full
Fig. 4. Experimentation scenario features of device can be found in [17].
IV. P ERFORMANCE EVALUATION
initiate a handoff process from WTP1 to WTP2. Here the In this section we describe the experimentation scenario
controller proceeds to migrate LVAP1 from WTP1 to WTP2. and analyze the obtained results. The goal is to compare the
The above is possible thanks to the fact that BSSID is unique handoff behavior in a traditional IEEE 802.11 network against
for each LVAP. This migration is performed without any link a SDN IEEE 802.11 network and to show the SDN impact in
or network layer messages or MU modification. the handoff process.
The experiments are performed under the following condi-
III. I MPLEMENTATION ON REAL NETWORK DEVICES USING tions:
SDN
• They were performed at the GITA research group labo-
The main contribution of this work is the implementation ratory of the University of Antioquia.
of a working framework for wireless networks using SDN • They were carried out within office hours, when there are
and the implementation of this on real network devices of different types of traffic in the wireless environment.
high performance and easy consecution in Colombia. This • There was a noisy wireless environment. Around 8 to 10
will allow to have a testing environment for the design of IEEE 802.11 co-located networks were present during the
new handoff techniques based on SDN. The objective of the tests.
real implementation is to compare the handoff process in a • No authentication credentials were used.
traditional scenario against a SDN scenario. • UDP traffic was generated every 0.5 seconds with the
We selected the EmPOWER framework [9], [10] due to the iperf tool [18].
following features: • Both APs operated in the same channel.
• Is an open and free code framework. • Each experiment was run during 100 seconds.
• Uses python as programming language. • The experiments were repeated 50 times.
• Is not necessary that switch used to interconnect the APs The measured variables were the following:
is an openflow switch. • Throughput in terms of bandwidth (200 samples per
• It allows the implementation of events oriented network
experiment).
applications. • Handoff delay (for this variable, 4 samples are obtained
• It is a framework in constant updating and it is maintained
per experiment, this means 4 handoff process per exper-
by its developers. iment).
• The hardware needed is of easy acquirement in Colombia
at low cost. A. Experimentation Scenario
EmPOWER can be considered as a Network Operating Sys- Fig. 4 presents the experimentation scenario to assess the
tem - NOS. EmPOWER exposes interfaces to the applications SDN impact in the handoff process. Traditional and SDN
to communicate with data plane devices without performing scenarios are the same, the only difference is the controller
important hardware modifications. Currently EmPOWER has introduced in SDN. The elements used in the experimentation
support for SDN wired networks, cellular and IEEE 802.11 scenario are summarized in table I.
networks. Detailed information of the framework can be found Since GITA’s lab is small (around 9.8*3.8 meters) it was not
in [11]. For the implementation of the SDN network in real possible to generate the handoff process, even if APs are set
network devices using Empower, one must keep in mind the up to the minimun transmision power. The handoff process
requirements of the controller [12] and the WTPs [13]. had to be emulated leaving the MU in a fixed equidistant
Table I. Network devices features
Device Reference O.S CPU Memory Wired card Wireless card
Intel(R) Core(TM)
DHCP LENOVO Ubuntu 4 GB
i5-3550 Realtek RTL8111/8168/8411 NA
Server MT 359 14.04.5 LTS DDR3
CPU @ 3.30GHz
HP 2920-24G Tri Core ARM1176 512 MB
Switch NA 24 RJ-45 10/100/1000 ports NA
(J9726A) @ 625 MHz SDRAM
TP-Link OpenWRT Qualcomm Atheros Atheros AR8327 Qualcomm Atheros
AP1, AP2 Archer C7 15.05 QCA9558 128 MB 4 10/100/1000Mbps QCA9558 + QCA9880
AC1750 2.0 Chaos Calmer CPU @ 720Mhz LAN Ports BR4A
Intel(R) Core(TM)
UDP ASUS Ubuntu 8 GB
i5-6200U RTL8111/8168/8411 Qualcomm Atheros
Server X556UQ 16.04.3 LTS DDR4
CPU @ 2.30GHz
Intel(R) Core(TM)
Mobile ASUS Ubuntu 6 GB Qualcomm Atheros Qualcomm Atheros
i5-2450M
User K53E 17.04 DDR3 AR8151 v2.0 AR9285
CPU @ 2.50GHz

25 25
Throughput (Mbps)

Throughput (Mbps)
20 20

15 15

10 10

5 5

0 0
0 20 40 60 80 100 0 20 40 60 80 100
Experiment time (s) Experiment time (s)
(a) IEEE 802.11 traditional network (b) IEEE 802.11 SDN network

Fig. 5. Throughput behavior for both scenarios

Table II. Mean throughput in IEEE 802.11 traditional scenario


Experiment 1 2 3 4 5 6 7 8 9 10
Throughput
7.58 7.82 7.65 7.61 7.48 8.19 7.72 7.83 8.29 8.13
(Mbps)

Table III. Mean throughput in IEEE 802.11 SDN scenario


Experiment 1 2 3 4 5 6 7 8 9 10
Throughput
9.50 7.46 8.04 8.59 6.14 7.08 8.12 8.43 7.19 8.25
(Mbps)

position from the APs and performing some procedures in experiment in the traditional scenario. It can be observed that
APs. In the case of traditional networks, the AP’s wireless the throughput fall at 20, 40, 60 and 80 seconds. In each of
cards were alternately turned down via bash commands every these times a handoff process occurred in the MU between the
20 seconds forcing the MU to initiate the handoff process. two APs. The throughput abruptly falls as a consequence of the
For SDN networks the LVAP for the MU was moved from MU disconnection from the network, having a zero value. This
AP1 to AP2 and vice versa using python instructions from is the expected behavior since we generated an instantaneous
the controller every 20 seconds. The aforementioned strategies disconnection. Table II presents the mean throughput for 10
allow to generate an instantaneous disconnection. If the MU out of the 50 performed experiments. It can be observed
would move, the throughput measure would not be affected that there is no significant variation in the mean throughput
only by the handoff process, but by the distance too. between different experiments.
Throughout the experiment time, MU performs four hand- Even though the throughput is an useful indicator of the
offs at 20, 40, 60 and 80 seconds respectively where each telecommunications networks behavior, the handoff delay, i.e.,
handoff method (traditional and SDN) performs its respective the time elapsed between the MU disconnection/connection
operation. from the network, is a more interesting metric to analyze
the handoff process. The handoff delay allows to isolate
B. Performance Results completely the handoff process and analyze the method ef-
1) IEEE 802.11 traditional networks: Fig. 5a shows the fectiveness.
throughput behavior of the handoff process for one particular Fig. 6 presents the mean handoff delay for each of the
four handoff processes with a 95% confidence interval. The
measures are performed using the iw tool [19] and taking each
process separately, i.e., the first handoff (at 20 seconds) of the 2.0
50 experiments is averaged, the second handoff (40 seconds)

Handoff delay (s)


of the 50 experiments and so on. Our key observations are 1.5
summarized as follows: 1) All the mean handoffs delay were
greater to 1 second. Given that MU is disconnected from the
1.0
network during the handoff process, this implies at least 1
second of service disruption, 2) The handoff processes are not
symmetric. When the MU performs the handoff from AP1 to 0.5

AP2 and from AP2 to AP1, the handoff delay is not the same
and 3) The throughput could not be a good metric to analyze 0.0
1 2 3 4
the handoff process. The mean throughput is not significant Handoff number
affected for packets loss over the handoff process. This can
be explained as the total time that MU was disconnected to the Fig. 6. Handoff delay for IEEE 802.11 traditional scenario
network is around 6 seconds (12 packets) against 100 seconds
(200 packets) of the overall experiment.
2) IEEE 802.11 SDN networks: Fig. 5b shows the through- 1.2
put behavior of the handoff process for one particular exper-
iment in the SDN scenario. While the MU is performing the 1.0
handoff process at 20, 40, 60 and 80 seconds, the throughput

Handoff delay (ms)


does not present abruptly falls with zero value, since the MU 0.8

is always connected to the SDN network. The throughput


0.6
variations can be explained by the wireless noisy environment.
Table III presents the mean throughput for 10 of the 50 0.4
performed experiments. As in traditional scenario, it can be
observed that there is no significant variation on the mean 0.2

throughput between different experiments.


0.0
In this SDN scenario, the iw tool did not register any event 1 2 3 4
that shows a MU disconnection from the SDN network. This is Handoff number
consistent with the throughput behavior and it implies that the Fig. 7. Handoff delay for IEEE 802.11 SDN scenario
MU does not disconnect from the network when it performs
handoff using SDN throughout a LVAP migration, in other
words, the MUs performs a seamless handoff. Considering
the aforementioned reasons, a valid strategy to estimate the two methods: standard (traditional) handoff and SDN hand-
handoff delay is to measure the execution time that controller off via LVAP. We performed experiments over real network
takes in to migrate the LVAP. Fig 7 presents the mean handoff devices considering a traditional IEEE 802.11 and a SDN
delay for each of the four handoff processes with a 95% controlled WiFi network.
confidence interval. These measures were realized under the The performance of traditional handoff and SDN handoff
same considerations that in the traditional scenario. It can strategies were assessed throughout experimental data the
observed that the handoff delay in SDN is in the miliseconds following terms of throughput (Mbps) and handoff delay with
scale against one second or more in traditional networks. following key results: 1) SDN allows to perform seamless
Our key observations can be summarized as follows: 1) SDN handoff process in IEEE 802.11 networks, 2) The LVAP
allows to perform a seamless handoff, 2) SDN’s handoff is abstraction performs the handoff process without modifications
completely transparent and without any modifications in the in the MU and 3) For the tested scenarios, the handoff delay
MU, 3) SDN allows to perform a handoff network controlled in traditional networks was in the scale of seconds against
process. This could be used to design new handoff triggers, miliseconds in SDN.
e.g. traffic categories, load balancing among others and 4) the The methods have been analyzed considering experimental
handoff delay in SDN network is in the miliseconds scale scenarios where APs are operating in the same channel. Future
against one second or more in traditional networks. work will be dedicated to the study of new approaches that
enable LVAP migration to APs operating on different channels
V. C ONCLUSIONS or when the migration is performed to a busy resource. On
The intention of this work was to study the handoff the other hand, since LVAP abstraction enables a network
management process in IEEE 802.11 networks. This process controlled handoff, the possibility to create new triggers where
is ruled for three phases: discovery, re-association and re- network information can be considered (traffic categories, load
authentication and it is governed by the MU. We considered balancing) is opened. In particular, some works have been
explored the LVAP migration to APs operating in different
channels [20].
ACKNOWLEDGMENTS
This work has been partially funded by the COLCIENCIAS
project COL16-01-01 of the University of Antioquia (Colom-
bia).
R EFERENCES
[1] M. Yang, Y. Li, D. Jin, L. Zeng, X. Wu, and A. V. Vasilakos, “Software-
defined and virtualized future mobile and wireless networks: A survey,”
Mobile Networks and Applications, vol. 20, no. 1, pp. 4–18, 2015.
[2] Y. Fang and W. Ma, “Mobility management for wireless networks:
modeling and analysis,” Wireless communications systems and networks,
pp. 473–512, 2004.
[3] D. Kreutz, F. M. Ramos, P. E. Verissimo, C. E. Rothenberg, S. Azodol-
molky, and S. Uhlig, “Software-defined networking: A comprehensive
survey,” Proceedings of the IEEE, vol. 103, no. 1, pp. 14–76, 2015.
[4] L. Suresh, J. Schulz-Zander, R. Merz, A. Feldmann, and T. Vazao,
“Towards programmable enterprise wlans with odin,” in Proceedings of
the first workshop on Hot topics in software defined networks. ACM,
2012, pp. 115–120.
[5] N. McKeown, T. Anderson, H. Balakrishnan, G. Parulkar, and Peterson,
“Openflow: enabling innovation in campus networks,” ACM SIGCOMM
Computer Communication Review, vol. 38, no. 2, pp. 69–74, 2008.
[6] I. . W. Group et al., “Ieee standard for information technology-
telecommunications and information exchange between systems-local
and metropolitan area networks-specific requirements part 11: Wireless
lan medium access control (mac) and physical layer (phy) specifica-
tions,” IEEE Std, vol. 802, no. 11, 2010.
[7] N. Cardona, J. F. Botero, and D. Ospina, “Handoff management for
smart access points in ieee 802.11 networks,” in Communications and
Computing (COLCOM), 2017 IEEE Colombian Conference on. IEEE,
2017, pp. 1–6.
[8] J. Schulz-Zander, P. L. Suresh, N. Sarrar, A. Feldmann, T. Hühn, and
R. Merz, “Programmatic orchestration of wifi networks.” in USENIX
Annual Technical Conference, 2014, pp. 347–358.
[9] R. Riggio, T. Rasheed, and F. Granelli, “Empower: A testbed for
network function virtualization research and experimentation,” in Future
Networks and Services (SDN4FNS), 2013 IEEE SDN for. IEEE, 2013,
pp. 1–5.
[10] R. Riggio, M. K. Marina, J. Schulz-Zander, S. Kuklinski, and
T. Rasheed, “Programming abstractions for software-defined wireless
networks,” IEEE Transactions on Network and Service Management,
vol. 12, no. 2, pp. 146–162, 2015.
[11] “Python wifi api documentation,” https://github.com/5g-empower/5g-
empower.github.io/wiki/Python-Wi-Fi-API-documentation.,
accessed:14.01.2018.
[12] “Setting up the empower controller,” https://github.com/5g-empower/5g-
empower.github.io/wiki/Setting-up-the-EmPOWER-Contoller, 2018, ac-
cessed:14.01.2018.
[13] “Setting up the empower wtp,” https://github.com/5g-
empower/5g-empower.github.io/wiki/Setting-up-the-WTP, 2018,
accessed:14.01.2018.
[14] “Openwrt,” https://openwrt.org/, 2018, accessed:14.01.2018.
[15] “Radiotap,” http://www.radiotap.org/, 2018, accessed:14.01.2018.
[16] “Bssid mask on atheros cards,” https://github.com/torvalds/linux/blob/
5924bbecd0267d87c24110cbe2041b5075173a25/drivers/net/wireless/ath/hw.c,
2013, accessed:14.01.2018.
[17] “Techdata: Tp-link archer c7 ac 1750 2.0,” https://goo.gl/2B1a9Q, 2018,
accessed:14.01.2018.
[18] “Iperf,” https://github.com/esnet/iperf, 2014, accessed:14.01.2018.
[19] “About iw,” https://wireless.wiki.kernel.org/en/users/documentation/iw,
2017, accessed:14.01.2018.
[20] L. Sequeira, J. L. de la Cruz, J. Ruiz-Mas, J. Saldana, J. Fernandez-
Navajas, and J. Almodovar, “Building an sdn enterprise wlan based on
virtual aps,” IEEE Communications Letters, vol. 21, no. 2, pp. 374–377,
2017.

Вам также может понравиться