Content uploaded by Jesus Alcober
Author content
All content in this area was uploaded by Jesus Alcober
Content may be subject to copyright.
Content uploaded by Jesus Alcober
Author content
All content in this area was uploaded by Jesus Alcober
Content may be subject to copyright.
Analysis of Control and Multimedia Real-Time Traffic over SIP and RTP on
802.11n Wireless Links for Utilities Networks
Salvador Santonja-Climent, David Todoli-Ferrandis,
Teresa Albero-Albero, Victor-M. Sempere-Pay´a,
Javier Silvestre-Blanes
ITI. Universitat Polit`ecnica de Valencia.
{ss,dt,talbero,vsempere,jsilves}@iti.upv.es
Jesus Alcober
Telematics Engineering Department
Universitat Politecnica de Catalunya.
jesus.alcober@upc.edu
Abstract
Urban facility interconnection networks require a ro-
bustness and reliability usually found in public networks
or private cable and radio networks in licensed bands.
Both factors mean an increase in costs, which also in-
crease as the size of the network increases. The recent
rise in use of wireless technologies in open wavebands
has attracted the interest of the industry in the spread of
these types of links, even though they have a lower level of
robustness and reliability, which must be improved using
communication mechanisms and protocols. This aspect
directly affects real time applications, such as VoIP and
video-streaming, and determines the viability of these sys-
tems in real installations. This paper evaluates the per-
formance of control and real time multimedia traffic in the
5 GHz band through the use of the SIP (Session Initiation
Protocol) and RTP (Real-Time Transport Protocol) proto-
cols, in a laboratory testbed.
1. Introduction
Waste water networks in metropolitan areas operate
over a network of distributed stations which fulfill a dou-
ble function of monitoring and actions, sending environ-
mental and water quality data to the central station, and re-
ceiving orders to be executed for sluices and pumps. The
traffic generated by these stations can be grouped into:
control and telemetry traffic, voice traffic (between opera-
tors and central) and video traffic (coming from cameras
that monitor water levels, obstructions or collectors sta-
tus). In general, the communication from the central to
the remote stations is carried out using heterogeneous net-
works, which combine the coverage and performance of
public networks (both wired and wireless) and private net-
works. The use of private networks is interesting in those
segments where they have spread enough to be technically
and economically viable, which will mean a considerable
saving in costs of use and greater control of the network.
Industrial networks giving support to these types of ins-
tallations are characterized by high reliability, adaptabil-
ity and scalability. However, the use of fieldbuses and
other wired technology are limited to local use, which
has sparked interest in the use of communication tech-
nology with wider coverage for interconnection in urban
areas. It is usual in this context to see the use of wireless
technology over licensed bands, which allows the use of
an interference-free channel as well as high transmission
power, which in turn means greater range and a higher
level of signal/noise relationship in the receivers, giving
greater reliability. However, the current trend is to use
open bands, such as 868 MHz, 2.4 GHz or 5 GHz. The
use of these bands offers significant benefits in terms of
installation and using costs, as there is an extremely active
market which is generating a huge quantity of new tech-
nology and products. However, there are certain disad-
vantages which must be considered, such as the reduction
in transmission power limits, which translates into a lower
range and weaker signal/noise relationship. Another dis-
advantage is that the band is shared with other users so on
occasions may be saturated, as 2.4 GHz band. Industrial
networks require high reliability and availability in com-
munications, which means it is necessary to study mecha-
nisms which can provide better guarantees for the traffic
in these types of technologies. In this paper, the behavior
of this three types of traffic necessary for the interconnec-
tion of remote stations of a real metropolitan waste water
system using 802.11n [2] radio technology and the SIP
[4] and RTP [5] session and transport protocols, is studied
and evaluated. The experiments were carried out on a lab-
oratory testbed, in order to obtain greater control over the
tests, and further experiments will be conducted in a real
exterior facility.
The need for new services in Valencia’s waste water
system, such as voice and video, has led to a gradual
increase of means of access to public networks, installed
by tele-operators to give service to the stations, which
involves a monthly cost as well as the externalization
of control of the network. With the objective of using
a private network and reducing costs, we have used a
laboratory testbed to recreate 1 and 2 hop links with
978-1-4244-6849-2/10/$26.00 ©2010 IEEE
802.11n, and carried out an initial evaluation of the
services offered over this real network. Fig. 1 shows an
aerial photograph of the Central Control Station (CCS)
and the nearby installations with which it must maintain
fixed links. In this image, we can see 1 hop links (stations
1 and 2) and 2 hop links (stations 3, 4 or 5). The use of the
5 GHz band means there are no costs involved, and there
is less saturation than with the 2.4 GHz band, as well as
a greater quantity of available channels, making WiMax
and 802.11n good candidates for these applications [6].
2. Scenario
The testbed is made up of 3 Linksys WRT602N routers,
with channels of 40 MHz bandwidth and a physical dis-
tance of 5 meters between each. Using virtual card confi-
guration and bridge mode, they were connected by 2 hops
at the data link layer level, obtaining a high capacity trunk
with throughputs of up to 80 Mbps from end to end. In
order to create a situation as close to real as possible, the
SIP server and a client are connected to one end of the
trunk, corresponding to the central control station, and the
second client is located in the intermediate router or at the
other end of the trunk, in order to carry out tests of 1 and
2 hops via remote station (Fig. 2). The server uses Open-
SIPS 1.6.2 and Asterisk 1.2.21.2 to offer these services,
and clients use free softphone X-Lite 2.0 for Linux envi-
ronments and 4.0 for Windows.
3. Technologies used
The 802.11n standard was ratified by the IEEE orga-
nization on 09/11/2009, with a speed of up to 600 Mbps
in the physical layer. Its range and quality are improv-
ing thanks to new MIMO (Multiple Input Output) tech-
nology, which allows it to use several channels simultane-
ously to send and receive data, thanks to the incorporation
of several antennas. The use of OFDM modulation offers
greater resistance to interference in the channel. 802.11n
can work in two frequency bands: 2.4 GHz and 5 GHz,
and at 20 or 40 MHz. Thanks to techniques to improve in-
terference resistance, transmission speed and bandwidth,
this new standard is becoming a real alternative, both eco-
nomically and in terms of performance, to WiMaX for
links of medium distance between urban installations.
Session Initiation Protocol, developed by the working
group MMUSIC (Multiparty MUltimedia SessIon Con-
Figure 1. Purification Water Facilities
trol) of the IETF (Internet Engineering Task Force), is
a signaling protocol for interactive multimedia sessions.
SIP is a protocol oriented toward the internet, and which
can deal with mobile users. It has some extremely interest-
ing characteristics, such as low processing needs, authen-
tication and encryption systems, modularity, extendibil-
ity and simplicity of code and network scalability. SIP
is the protocol that deals with establishing and finishing
the communication, as well as negotiating the necessary
parameters for this communication, such as the compres-
sion algorithms used, sampling frequencies and multime-
dia characteristics, pricing and information security. SDP
(Session Description Protocol) is used to describe the con-
tent of the session, and RTP as the carrier for the voice
and video of the session. The choice of SIP + RTP over
UDP for the transfer of multimedia content offers a po-
werful call routing and control system between clients,
and ensures a constant flow and organized delivery thanks
to temporal marking techniques for the synchronization
of the connection and reorganization of packets received
using RTCP (RTP Control Protocol).
4. Evaluation
4.1. Control Traffic
Control traffic is made up of short sequences which
carry readings from sensors or initiate actions over the
PLCs (Programmable Logic Controller) of the station. In
our case the station’s largest frame would be 592 bytes of
payload, this can be transmitted in a single Ethernet frame.
Evaluation must consider the delay between transmission
and reception (latency) and the percentage of losses. This
can be done using the pingSIP tool [1] which sends a SIP
message of type OPTION which circulates at the session
level between origin and destination and diagnoses the
proper functioning of the end to end link and across all
levels involved. In contrast to ping, pingSIP does not suf-
fer from blocks in routers with NAT and reaches directly
the destination device and requires greater processing in
the level of layer in which it is operating. Fig. 3 shows
the results from 300 pingSIP tests for one and two hops,
where a loss percentage of 0% was obtained in both cases.
For one hop, 77% of the values are concentrated at around
2.88 ms, with an arithmetic average of 5.79 ms (marked
in the figure with a dotted line) and a typical deviation of
5.73 ms. For two hops, the latency was concentrated at
around 3.05 ms, the arithmetic average was 6.52 ms and
the typical deviation was 7.49 ms. These latency values, in
the order of a few milliseconds, are negligible for the tar-
Figure 2. Testbed
2
Figure 3. PingSIP results
Figure 4. Audio Codecs throughput
get control applications, which demand latencies of sev-
eral seconds, and so the system fulfilled all expectations.
4.2. Voice Traffic
Voice traffic needs certain parameters in order for the
conversation to be fluent. The recommended maximum
thresholds are latencies of 150 ms (according to ITU rec-
ommendation G.114), 100 ms of jitter and 1% losses. The
codec used is critical in this type of application as, due to
the digitalization and compression of voice used [3], the
frequency of transmission of frames as well as their size
varies. The codec itself therefore generates a fixed delay
in frame transmission due to the need to fill it with part
of the conversation. The codecs evaluated and their cha-
racteristics are shown in Table 1. Figs. 4 show that codec
G.711u uses more bandwidth, due to a higher bitrate for
a better quality, with less compression complexity. How-
ever, the other codecs use less bandwidth and tolerate net-
works with less capacity. GSM is a codec widely used
in mobile telephones, whose voice quality is low com-
pared to more modern codecs. In iLBC (internet Low
Bitrate Codec), the loss or delay of packets is balanced
out through degradation of voice quality. iLBC is a good
codec for environments with high losses, but with lower
performance in controlled environments or those with low
losses. SPEEX maintains a variable codification which
Table 1. Audio codecs under evaluation
Codec bitrate Frame Size
G.711u-a 64kbps 20ms
GSM 13 kbps 20ms
iLBC 13.33 kbps 30 ms
SPEEX 2.15-24.6 kbos 40ms
Sampling frequency at 8 KHz in all cases.
Figure 5. Jitter
reduces the bitrate when necessary. It is a codec oriented
toward VoIP, and not toward telephones, for which reason
it is optimized for a high quality voice service with low
bitrates. Fig. 5 show the jitter for the same audio transmis-
sion for 2 minutes for each of the codecs mentioned. As
can be seen, the jitter remains constant for one as well as
two hops. For G711u, the jitter remains at around 0.25ms
for one hop and 0.48 ms for two hops. In GSM, it re-
mains at around 0.28 ms for one hop and 0.31 for two
hops. For iLBC it is 0.41 for one hop, and 0.43 for two
hops, and for SPEEX 0.39 ms for one hop, and 0.42 for
two. The variations in this harmonic means are negligi-
ble, with maximum peaks of 20 ms for two hops for iLBC
and SPEEX. An important detail to consider is the quan-
tity of packets generated in each codec, for the same time
period. With G.711u and GSM we can see 6000 pack-
ets received, which corresponds to the 120 seconds trans-
mission time between 20 ms per frame. With iLBC 4000
frames were generated, each containing 30 ms, and with
SPX 3000 frames, each containing 40 ms. Table 2 shows
the maximum separation between consecutive packets, the
average and maximum jitter and the packets loss. From
one to two hops, there is an increase in the average and
maximum jitter, but this value still remains well below the
thresholds for VoIP. The loss of packets does not exceed
0.06% in any of the cases, which is also well below the
limit of 1%.
Table 2. Audio results summary
1 hop G.711u-a GSM iLBC SPEEX
Max δ47.98 139.79 149 148
Max Jitter 4.24 12.33 12.3 12.07
Mean Jitter 0.3 0.6 0.66 0.62
Packets Lost 0.03% 0.02% 0.05% 0.06%
2 hops G.711u-a GSM iLBC SPEEX
Max δ123.37 150.63 152 148
Max Jitter 18.02 16.66 20.26 20.63
Mean Jitter 1.3 0.79 0.86 0.82
Packets Lost 0.03% 0.03% 0.00% 0.03%
all datas in ms, except lost packets.
3
Figure 6. Jitter/bitrare relationship in video
streaming
4.3. Video Traffic
Recommended quality parameters are 4 seconds of la-
tency, a jitter depending on the buffering of the system
and a loss rate of 5% [7]. Minimizing the need for buffer-
ing in the system improves the streaming delay, making
the transmission and visualization of the video take place
in practically real time. Asterisk supports codecs H261,
H.263, H.263+ and H.264. This last one is included in
the latest version of Asterisk, but does not have as much
compatibility with clients software, so H.263+ was cho-
sen. This codec works with most video-conference ap-
plications, offers high image quality with a high level
of compression, and has significant improvements in ef-
ficiency and robustness over its predecessors. The tests
were carried out using different codification rates and the
same 2 minutes video sequence. Fig. 6 shows the test re-
sults. Even for transmissions 1Mbps, the jitter remains at
around 6.7 ms, and the bursts remain below 30 ms, which
is extremely low. Carrying out more exhaustive tests using
transmission rates of 1Mbps, we obtained the instant jitter
that can be seen in Fig. 7 for one and two hops. With one
hop, the jitter stays around 3.35 ms, with a typical devia-
tion of 2.83 ms. With two hops, the harmonic average is
3.61 ms with a typical deviation of 2.94 ms. The loss rate
in both cases is 0%.
Figure 7. Jitter
5. Conclusions and Future Work
The evaluation of control, voice and video traffic in an
802.11n trunk of one and two hops shows optimum values
in all required quality parameters for these types of appli-
cations. The differences arising from increasing the trunk
from one to two hops are minimal in terms of average jitter
and losses, affecting mainly the latency and the through-
put. The use of different codecs allows better adaptation
to the capacities of the links, using lower bitrates for links
with lower throughput, but at the cost of lower quality, al-
though the tests carried out with video-streaming showed
a constant flow, without errors, at rates of 1 Mbps, offering
a high image quality, without cuts or failures in codifica-
tion, all of which results in a high quality user experience.
The evaluation of these parameters in the laboratory al-
lows greater control of the testbed, although in all proba-
bility the efficiency of a real installation would be slightly
reduced due to external factors, such as an increase in dis-
tances, obstacles, interference, climatic factors and indus-
trial noise. Once the viability of this type of traffic over
802.11n links has been thoroughly checked, it will then be
necessary to carry out tests in a real environment, in order
to obtain comprehensive conclusions on the performance
of this solution for low cost interconnection in urban fa-
cilities.
Acknowledgement
This work was supported by CICYT grant TSI2007-
66637-C02-01/02 and PET2007 0316, which are partially
funded by FEDER
References
[1] A. G´omez-Extremera. Implementar modulo de QoS para
VoIP en SIP. Final Thesis Report, Sept. 2006.
[2] IEEE 802.11 Working Group. 802.11n-2009 IEEE Standard
for Information technology. Amendment 5: Enhancements
for Higher Throughput.
[3] M. Karam and F. Tobagi. Analysis of the delay and jitter
of voice traffic over the internet. Proceedings of Infocom,
2001.
[4] J. Rosenberg, H. Schulzrinne, J. A. Camarillo, J. Peterson,
R. H. Sparks, and E. Schooler. SIP: Session Initiation Pro-
tocol. IETF RFC 3261, Jun. 2002.
[5] H. Schulzrinne, S. Casner, R. Frederick, and V. Jacobson.
RTP A Transport Protocol for Real-Time Applications.
IETF RFC 1889, Jan. 1996.
[6] J. Silvestre, V. Sempere, and T. Albero. Wireless Metropoli-
tan Area Networks for Telemonitoring Applications. 7th
IFAC Int. Conf. on Fieldbuses, Networks in Industrial and
Embedded Systems, Nov. 2007.
[7] C. Szigeti, T. and Hattingh. End-to-End QoS Network De-
sign: Quality of service in LANs,WANs and VPNs. Cisco
Press, Nov. 2004.
4
