Available online at www.sciencedirect.com
ScienceDirect
Procedia Computer Science 104 (2017) 493 – 500
ICTE 2016, December 2016, Riga, Latvia
Evaluation of Wi-Fi and LTE Integrated Channel Performance with
Different Hardware Implementation for Moving Objects
Arnis Ancansa,*, Nikolajs Bogdanovsa, Ernests Petersonsa, Guntis Ancansa,
Aleksandrs Umanskisa, Vishnevskiy Vladimirb
b
a
Riga Technical University, Azenes St. 12-317, Riga, LV-1048, Latvia
V.A.Trapeznikov Institute of Control Sciences of Russian Academy of Sciences, Profsoyuznaya str. 65, Moscow, 117997, Russia
Abstract
Combining Wi-Fi and LTE data transmission technologies we can obtain an integrated solution that provides full end-to-end data
transmission channel. This article compares goodputs of two LTE routers, which connect multiple WLAN access points
(including Wi-Fi access controller) with the remote FTP server. Handover impact on data channel goodput is measured between
wireless access points. Experimental data obtained by taking measurements under field conditions using real hardware, which
imitates realistic two rank heterogeneous wireless network. The obtained results are very valuable, as they can be used as source
data to develop mathematical network models as well as to tune network simulation software. The article also contains a
mathematical model, which allows analyzing the impact on the data channel goodput at several competing vehicles which try to
gain access to data channel via wireless access points. Experimental data is used in mathematical model. Obtained results
supplement previous research and in this case the results show hardware’s impact on channel goodput.
©
Published
by by
Elsevier
B.V.B.V.
This is an open access article under the CC BY-NC-ND license
© 2017
2016The
TheAuthors.
Authors.
Published
Elsevier
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review
under
responsibility
of
organizing
committee
the scientific
committee
of the international
conference;
ICTE 2016.
Peer-review under responsibility of organizing committee
of theofscientific
committee
of the international
conference;
ICTE 2016
Keywords: Heterogeneous network; Handover; Goodput; Wi-Fi; LTE
* Corresponding author. Tel.: +371-28331178.
E-mail address: arnis.ancans@rtu.lv
1877-0509 © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of organizing committee of the scientific committee of the international conference; ICTE 2016
doi:10.1016/j.procs.2017.01.164
494
Arnis Ancans et al. / Procedia Computer Science 104 (2017) 493 – 500
1. Introduction
Two major standardized technologies are currently being considered for Vehicle-to-Everything (V2X)
communications in automotive applications: IEEE 802.11p and 3GPP Long Term Evolution (LTE) with Proximity
Services (ProSe). However, these technologies have their drawbacks. For example, IEEE 802.11p cannot be used to
transfer large amounts of data, such as video, management and security functionalities of the vehicle (e.g.,
Certificate Revocation List (CRL) distribution) and the transfer of necessary data to roadside units (RSUs)
infrastructure and further to control centres. And also this standard solution cannot be used to access the Internet.
We could use existing cellular networks (e.g. LTE), but it may be inaccessible due to weak or non-existent coverage
in such places like tunnels, underground parking, rural area, mountainous terrain etc., all of which are important
places where road safety and other mentioned services must be ensured. Therefore, in order to address these issues,
authors offer to use an IEEE 802.11 based infrastructure mainly in places with no broadband mobile network
coverage. We also assume that the use of Wi-Fi infrastructure in such problematic areas is economically beneficial.
The integrated idea of ubiquitous 5G V2X network presence for vehicles and strict requirements for security
mechanisms to prevent unauthorized access to vehicles and related personal data, which are demanding against the
network delay, justifies position of authors even more. We can also mention the need in near future to transfer data
of sensors placed in the vehicle, as well as transport infrastructure data to 5G networks, which will provide timely
supply of data to compute and storage centers1. This confirms that previous and upcoming research in directions of
both Wi-Fi and LTE technology for moving objects are justified and necessary.
The main goals of the paper are to evaluate performance (goodput) of IEEE 802.11n2 and LTE integrated channel
for two rank heterogeneous wireless networks using different LTE channel access equipment and to evaluate impact
of handover (inter-AP) on data channel goodput between wireless access points.
In computer networks, goodput is the application-level throughput (i.e. the number of useful information bits
delivered by the network to a certain destination per unit of time). The amount of data considered excludes protocol
overhead bits as well as retransmitted data packets. This is related to the amount of time from the first bit of the first
packet is being sent (or delivered) until the last bit of the last packet is delivered.
In the Wi-Fi world, handoff, handover and roaming, are all the same, referring to final users moving between
different networks with or without supporting IP session continuity. With the same (unchanged) IP address, the IP
session continuity can be achieved3.
To perform the necessary measurements within this study a test-bed were created with further performance
evaluation. Experiments were carried out by real equipment in field environment with real road conditions. In first
case we used professional LTE router from Cisco and in second case - home type Huawei LTE router.
The results of this work supplement those reported in4,5,6,7 as in mentioned studies they are missing the review of
influence on communication channel performance depending on used LTE access hardware.
The rest of this article is organised as follows. The second chapter is devoted to the system description. The third
chapter describes the methodology and scenario of measurements. The fourth chapter is devoted to mathematical
models for heterogeneous network. Third and fourth chapters also holds the discussion on obtained results. Finally,
we came up with some final conclusions.
2. Description of system
The physical realization (test-bed, see Fig. 1) of communication for the transmission data from the vehicle to the
remote server and back is the wireless network. At the first stage the data are transmitted from the mobile object to
the nearest access point (AP) according to the IEEE 802.11n standard. Further, from the AP the data are transmitted
to the remote server by the channel according to the LTE.
Test-bed consisted of LTE (4G) router (Cisco 819G-4G-G-K9 or Huawei E5186), Wi-Fi access controller (HP
MSM720), 3 x Wi-Fi access points (HP MSM460), moving object (vehicle) equipped with powerful laptop (i53360M 2.8 GHz CPU, 4 GB RAM, Wireless LAN adapter Intel Centrino Advanced-N 6205) and remote FTP server
(E5335 2 GHz CPU, 4 GB RAM). Remote server and laptop were equipped with IxChariot software.
Arnis Ancans et al. / Procedia Computer Science 104 (2017) 493 – 500
495
Fig. 1. Network topology for test-bed.
Technical specifications of used LTE router hardware is given in Table 1. LTE download (DL) / upload (UP)
speeds depend on specific carrier channel bandwidth and carrier LTE network provisioning.
Table 1. Technical specifications of used hardware.
Parameter
Cisco 819G-4G-G-K9
Huawei E5186
Mobile wireless
technologies supported
LTE/UMTS/HSPA+/DCHSPA+/Three-band/
EDGE/GPRS/GSM
LTE/UMTS/EDGE/GPRS
/GSM
Frequency bands, MHz
800/900/1800/2100/2600
800/900/1800/210/2600
DL/UP speeds, Mbps
LTE CAT3:100/50
LTE CAT6:300/50
Antenna
Antenna pattern: omnidirectional
Antenna pattern: omnidirectional
Gain: 2 dBi
Gain: 5 dBi
Polarization: vertical
Polarization: vertical
During measurements Huawei LTE router’s RSSI (Received Signal Strength Indication) was -64 dBm and Cisco
LTE routers -60 dBm.
3. Methodology and scenario of measurements
Two different vendor’s LTE routers were used to evaluate possible impact on integrated Wi-Fi and LTE channel
goodput. Each LTE router connects Wi-Fi access controller and three WLAN access points with the remote FTP
server. Measurements were taken with one vehicle at different moving velocities: 20, 50 and 90 km/h, and also at
steady position.
Fig. 2 show comparison of Cisco LTE router and Huawei LTE router obtained results using only one moving
object – vehicle with velocity 20 km/h.
By increasing movement velocity, the transmission download speed falls. At a speed of 20 km/h the handover
between access points (AP) is carried out properly as shown with traffic downs. Connection becomes unstable at
higher speeds.
496
Arnis Ancans et al. / Procedia Computer Science 104 (2017) 493 – 500
Fig. 2. Comparison of two-stage goodput with LTE and 802.11n at the speeds of 20 km/h.
Standard client workstation configuration did not allow to measure download speed at 90 km/h, but switching
default Roaming Aggressiveness parameters from Medium to Highest did make it happen.
In Fig. 3a we can see active data transfer time in each access point (AP) zone. In Fig. 3b is presented the
assessment of switching time (handover) between access points.
a
b
Fig. 3. (a) Time of active data transmission in AP zone; (b) Handover time between APs.
As the Fig. 2 and Fig. 3 show, the measurements taken with various LTE devices gave different performance
results. The difference in performance measurements makes 10–15%, and, in general is related to the transition
process, when traffic transmission downs are observed as hollows.
Table 2. Summary of results.
Speed,
km/h
LTE router
20
Cisco LTE
50
90
Data transmission time in
AP zone (s)
Handover time
between APs (s)
1
2
3
1–2
2–3
20.21
8.94
12.31
7.25
4.97
Huawei LTE
14.22
10.10
12.04
10.82
8.93
Cisco LTE
7.73
7.14
4.03
1.52
1.59
Huawei LTE
7.89
2.74
4.21
2.72
3.95
Cisco LTE
3.26
0.74
0.72
5.89
2.11
Huawei LTE
2.87
1.29
1.24
5.73
1.55
497
Arnis Ancans et al. / Procedia Computer Science 104 (2017) 493 – 500
Summary of obtained experimental results is given in Table 2 where we can see active data transfer time in each
access point zone. For AP-1 data transmission zone is 1, for AP-2 zone is 2 and for AP-3 zone 3. This table also
provides a summary on handover time when mobile client travels between two pairs of access points which are
denoted as 1–2 and 2–3 zones.
Each hardware brings its own settlement coefficient. It was observed that by sing different LTE router hardware,
the switching time between APs were different. It is related to the transition processes, which results in the changes
of overall connection creation time.
It is assumed that different LTE router devices have different interference resistance and it may affect
performance measurements.
4. Mathematical models for heterogeneous network
Until this we described experiments with one moving vehicle with many access points equipped along the road.
This research is necessary in future to estimate two stage wireless transport system in real environments. In real
world many vehicles move by access points and some of them can send requests to AP. In result, clients can share
throughputs of access points and common channel from concentrator to server. Therefore overall goodput of the
system can be changed to ideal – only one moving vehicle.
How can one client evaluate system goodput with many clients being on road – these are major questions for
future researches. As an answer to this question this work suggests mathematical models for heterogeneous network.
To assess the model parameters we believe that the time used for data handling in nodes submits to exponential
distributions.
Based on network model described in second chapter and which was used in experimental stage, such system can
be represented by the two-stage network model, as it is shown in Fig. 4.
Fig. 4. Two-stage vehicular network model.
The route of data transfer goes from the zero node (mobile client) to the first node (access point) and further to
the second one (remote server), if the file transfer is considered from the vehicle to remote server over LTE channel.
The ACK (acknowledgment) confirmation of packet flow is transferred from server. In this case the average time of
transmission will differ: more time is spent on the transmission of the data packets, which is denoted as E(t i ) . ACK
transmission will take less time; let’s mark it as E(t 0 ) . Then the average time of data handling (data processing) in
the first node will be:
E(t1 ) =
E(t i ) + E(t 0 )
.
2
If ACK confirms each uploaded packet, the processing intensity of the zero node will be:
(1)
498
Arnis Ancans et al. / Procedia Computer Science 104 (2017) 493 – 500
ȝ0 =
1
.
E(t1 )
(2)
The first node imitates behaviour of a Wi-Fi network (wireless AP and wireless access controller) that works with
rated data speed. Parameters for the first node are received experimentally. The intensity of data handling is equal
to:
ȝ1 = H 1 ,
(3)
where ߝଵ – efficiency of 802.11n network received experimentally (see Fig. 2) and probabilities P10 = 0,999 and
P12 = 0,001 .
Parameter N defines the number of data transfer sources which compete for the distribution of node resources. In
our case this is the number of cars in AP coverage area. Then this three-nodal and two-stage goodput model can be
expressed in formula (10). Parameters and X in this formula are defined by values from (4). Data transmission rate
on wireless network of IEEE 802.11n standard depends on distance of the car from AP. It complicates the task as
ȝ1 will not be a constant value but vary. To assess the system it is necessary to know the number of cars in AP zone
and AP subzone. Subzones are understood as road sections with different distances from AP.
To compare two options of heterogeneous network goodput we use data from paper6, where 200 meter coverage
area of a APs is divided into 5 zones with 40 meters each. Request (car) processing time by each device of a road
stretch is random. It is distributed under the exponential law and is shown in Table 3.
Table 3. Cars in each zone.
N
E n1
E n2
E n3
E n4
E n5
10
6
1
1
1
1
20
15
2
1
1
1
Each subzone holds according value N i = E ni .
For further calculations following variables will be used:
X1 =
ȝ
ȝ0
; X 2 = aX 1 ; a = 1 P12 .
ȝ1 P10
ȝ2
(4)
Intensity for ȝ 2 :
1
ȝ2 = ,
t
t=
lp
Vf
(5)
,
(6)
where V f – actual speed of data transfer for LTE.
LTE technology is used for data transfer between LTE router and mobile network base station. For packet length
of lp=1500 bytes, actual speed obtained experimentally – V f .
Normalizing function G(N) obtained from the equality condition of unit sum probabilities p n0 , n1 , n2 , where ni
in vector n = n1 , n2 , n3 is the number of requests in i node. This work gives general expression for calculating
G(N):
499
Arnis Ancans et al. / Procedia Computer Science 104 (2017) 493 – 500
3
G( N )
n
¦3 Xi i ,
n i 1
(7)
where N – number of cars.
Application of Buzen’s algorithm7 for examined three-phase model leads to the following type of the specified
function:
G(N) =
1 N j
j+1
.
¦ X1 1 a
1 a j=0
(8)
We will determine performance Ș as a number of processed requests in a unit of time8,9,10. The complete task is
put through the input-output subsystem and a new one is instantly loaded. The output stream is equal to the input
stream and from the condition of stream balance we can conclude the following:
p^ni = 0`=
G(N) X i G(N 1 )
,
G(N)
(9)
where:
Ș = P10 ȝ1 ( 1 p^ni = 0`) .
(10)
Goodput for three-nodal closed network of Huawei and Cisco with P10=0.999 P12=0.001 for N=10 and N=20
shown in Fig. 5.
Fig. 5. Goodput for Huawei LTE and Cisco LTE networks.
From the Fig. 5 we can see that the network performance is influenced not only by probability models: P10 and
P12 , but also by the parameters of LTE router (Table II) that influences overall goodput of heterogeneous network.
Network with LTE router Cisco 819G has better goodput than Huawei E5186 network by 10–15%.
500
Arnis Ancans et al. / Procedia Computer Science 104 (2017) 493 – 500
5. Conclusion
According to received data, we can conclude that the performance of the Wi-Fi and LTE Integrated Channel
appears to be connected not only to the parameters of traffic movement, but also to parameters of used wireless
WAN (mobile broadband backhaul) router.
Subsequent studies will carry out experimental implementation of more complex models with more vehicles
involved. It will allow obtaining experimental measurement results that meet realistic traffic flow conditions. It is
also planned to start the research and find solutions to ensure optimal roaming (inter-AP handover) mechanism for
moving objects i.e. Wi-Fi AP clients who are on the move.
The acquired results of evaluation can be used by communication equipment manufacturers, car manufacturers,
mobile operators and other interested parties when planning next generation (5G) mobile services.
References
1.
5G Automotive Vision; 2015. Available: https://5g-ppp.eu/wp-content/uploads/2014/02/5G-PPP-White-Paper-on-AutomotiveVertical-Sectors.pdf.
2.
Ancans G, Stankevicius E, Bobrovs V, Ancans A. Analysis on Interference Impact of Wi-Fi on Digital Terrestrial Television
Broadcasting. International Journal of Interdisciplinary Telecommunications and Networking (IJITN). Vol. 8 (1); 2016. p. 35–44.
3.
Interworking Wi-Fi and Mobile Networks. Ruckus Wireless; 2013. Available: http://c541678.r78.cf2.rackcdn.com/wp/wpinterworking-wi-fi-and-mobile-networks.pdf.
4.
Tan WL, Lau WC, Yue OC, Tan HH. Analytical Models and Performance Evaluation of Drive-thru Internet Systems. IEEE Journal
on Selected Areas in Communications. Vol. 29 (1); 2011. p. 207–222.
5.
Bogdanovs N, Petersons E. Performance Evaluation of Heterogeneous Network for Two Vehicle Regime. In: Advances in Wireless
and Optical Communications (RTUWO 2015). Riga; 2015. p. 164–167.
6.
Ipatovs A, Bogdanovs N, Petersons E. Evaluation of the Rate of the Data Transfer between Vehicles and Base Stations in Wireless
Networks. Automatic Control and Computer Sciences. Vol. 46 (2); 2012. p. 76–82.
7.
Bogdanovs N, Petersons E, Jansons J. Two Layer Model for Performance Evolution of V2I Network. Electronics and Electrical
Engineering. Vol. 19 (3); 2013. p. 98–101.
8.
Garber NJ, Hoel LA. Traffic & Highway Engineering. Cengage Learning; 2009. 1230.
9.
Jain R, Smith JM. Modeling vehicular traffic flow using m/g/c/c state dependent queueing models. Transportation Science. Vol. 31;
1997. p. 324–336.
10. Gross D, Shortle J, Thompson J, Harris C. Fundamentals of Queuering Theory. Willy; 2008. 500.
Arnis Ancans received the bachelor’s degree of Engineering Science in Electrical
Engineering and master’s degree of Engineering Science in Telecommunications at Riga
Technical University in 2003 and in 2005 respectively. He is currently continuing his
research in Riga Technical University as a doctoral student in the field of wireless network
performance research. He is a senior network administrator at Latvenergo AS which is
engaged in generating electric and thermal energy and providing telecommunication and
information technology services. His main research interests are in wireless LAN with
moving objects, intelligent transportation systems, IP data network monitoring systems.
Contact him at arnis.ancans@rtu.lv.