See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/229008214

QoS Requirements of Network Applications on the Internet

Article · January 2004

DOI: 10.1145/1030194.1015475

CITATIONS READS

225 12,273

3 authors, including:

Toni R Farley
Arizona State University
24 PUBLICATIONS 771 CITATIONS

SEE PROFILE

All content following this page was uploaded by Toni R Farley on 30 May 2014.

The user has requested enhancement of the downloaded file.

Information ● Knowledge ● Systems Management 4 (2004) 55–76 55
IOS Press

QoS Requirements of Network Applications on the

Internet
Yan Chen, Toni Farley and Nong Ye*
Department of Industrial Engineering, Arizona State University, Tempe, AZ, USA

ABSTRACT: In this paper, we present quality of service (QoS) metrics for various network applications based on
human factors and technology attributes. The first term, human factors, addresses human perception of different
kinds of media, such as conventional text, audio and video. The second term, technology attributes, represents the
different technological aspects of these network applications, such as time-dependence and symmetry. Both of these
terms are key factors that lead to variations of requirements for QoS. Establishing these requirements is paramount
to providing QoS on computer networks and the Internet. With the metrics presented in this paper we can provide
the criteria necessary for such QoS assurance.

INTRODUCTION

Information services on the Internet come in varying forms, such as web browsing, e-mail, and
multimedia on demand. The main motivation behind the design of next-generation computer and
communications networks is providing universal and easy access to these various types of information
services on a single multi-service Internet. This means that all forms of communications (video, voice,
data and signaling), along with all types of services (from plain text web pages to multimedia
applications), are bonded in a single-service platform through Internet technology.
Media are transmitted on a network by bit streams. Each bit stream is represented by a flow of bits of
1s and 0s through the network. Although all media are transmitted in the same digital form, each
application generating a bit stream has an associated set of service characteristics. Regardless of the type
of media (picture or voice) there are different QoS requirements perceived from the user’s standpoint that
determine whether the user finds this type of service acceptable. Examples of user perception include
phone-to-phone delay and playing music online.
In general a phone-to-phone delay, from a user’s perspective, should be no more than 150 milliseconds
(Silveira et al, 1996) to allow for appropriate and easy understanding. Outside of this limit, the user will
become annoyed and find the service unacceptable. To support this kind of a call over the Internet, the
voice media need to be transmitted with a constant and reliable bit transmission rate, low jitter, and a low
error and loss rate.
User perception is also important when a user plays music online. If the variation of delay across the
Internet is too large the song sounds desultory, and the user stops listening. When a network cannot
provide the corresponding support necessary to make a service acceptable, the user feels that this kind of
service is senseless.
The Internet today is made up of a simple convergence of joining application media bit streams. This
convergence does not take into account the service characteristics associated with each type of
*
Corresponding author: Professor Nong Ye, Department of Industrial Engineering, Arizona State University, Box 875906,
Tempe, AZ 85287-5906, USA. Tel.: +1 480 965 7812; Fax: +1 480 965 8692; E-mail: [email protected].

1389-1995/04/$17.00 © 2004 – IOS Press and the authors. All rights reserved
56 Y. Chen et al. / QoS Requirements of Network Applications on the Internet

application. This means that the Internet does not currently provide any quality assurance.
Concise specification of service requirements is vital to realizing quality assurance on computer
networks. Since different applications have different service features, each application should specify its
explicit requirements to a computer network in order to achieve the desired QoS. If there are no
requirements given, the network will take for granted that any level of service is acceptable, and therefore
can provide any level of network support. The types of network support provided, such as bandwidth and
priority in a router’s queue, may lead to delay and jitter that can then render the service unacceptable.
Consequently, if we want to satisfy QoS of various applications in next-generation networks, the first step
is to identify the QoS metrics of each network application.
In this paper, we first analyze the influence of two key factors, human factors and technology
attributes, on QoS requirements. Next, we present network application classification based on the
application’s different service characteristics, resulting mainly from these two key factors. Finally, based
on existing literature, we systematically propose numerical measures, which can quantitatively represent
the QoS requirements of various applications with different service characteristics in Multi Service
Networks, along with the rationale behind these choices.

QOS MEASURES

In general, QoS has three attributes to measure the output performance of a process: timeliness,
precision and accuracy (Ye, 2002). Timeliness measures the time taken to produce the output of the
process. Precision measures the amount or quantity of the produced output. Accuracy measures the
correctness of the produced output, usually relating to the content of the output. Specific measures of the
three QoS attributes depend on the process of interest. Existing work on QoS of computer networks has
used the following QoS measures.
Response Time Expected by Users: The users’ expected response time is the time elapsed between
sending a request and the reception of the first response by the user.
Delay: The network transmit delay is the time elapsed between the emission of the first bit of a data
block by the transmitting end-system, and its reception by the receiving end-system.
Jitter: In transmission technology, jitter refers to the variation of delay generated by the transmission
equipment.
Data Rate: Data rate refers to the raw data rate of encoded multimedia data before transmission, that is,
the rate in which data are encoded.
Required Bandwidth: The required bandwidth is defined by the required data transfer rate, measured in
bits per second, of each specific application in telecommunication. This metric includes raw data and
overhead.
Loss Rate: The bit loss rate is the number of bits lost between two points in telecommunications after
transmission.
Error Rate: The bit error rate is the frequency of erroneous bits between two points in
telecommunication after transmission.

TECHNOLOGY ATTRIBUTES OF APPLICATIONS

Different applications can be characterized by certain technology attributes. Applications in one class
of technological attribute may not share the same QoS requirements as those in another class. The
technology attributes used to classify applications in this paper are defined in this section.
 Y. Chen et al. / QoS Requirements of Network Applications on the Internet 57

Time Dependence

Applications can be classified by their time dependency requirements. We classify all applications into
two time dependant categories: real time (RT) applications and non real time (NRT) applications.
Real time is defined by the Oxford Dictionary of Computing as: “A system in which the time at which
the output is produced is significant. This is usually because the input corresponds to some movement in
the physical world, and the output has to relate to that same movement. The lag (delay) from the input
time to output time must be sufficiently small for acceptable timeliness”.
RT applications can further be divided into soft real time and hard real time. The main difference
between soft and hard real time is that hard real time applications will fail should QoS requirements not
be met. In this paper we group all soft and hard real time applications as RT.
In RT applications, the network needs to deliver time-based information without changing its built-in
time properties. For adequate user satisfaction, we need to maintain more stringent delay and jitter
requirements for RT applications. The delay requirements must be strict in order to maintain system
timing. The jitter requirements are essential to transmitting data at a constant, reliable rate.
NRT applications are any applications that do not have stringent timing requirements. This type of
application does not fail if timeliness metrics are not met, nor does it require timing accuracies to be
considered acceptable. NRT applications, which do not have time-based sensitivity requirements, are
mostly concerned with delay.

Symmetry

An application’s symmetry property allows us to classify all applications into two categories:
symmetric applications and asymmetric applications.
In symmetric applications, requests and responses are comparable in terms of resource consumption.
Videophony is an example of a symmetric audio and video application. This application requires equal
resource consumption on both host machines. In this case, the request consumes an equivalent amount of
resources as the response.
In asymmetric applications, requests are considerably less resource consuming than responses. An
example of an asymmetric application is Video-on-Demand (VoD). This application is highly asymmetric
since it consumes much more resources in the VoD Server (response) than in client machine (request).
The grades of symmetry in multimedia services result in different grades of interactivity, which in turn
lead to different response time and delay requirements. For example, Videophony requires an almost
instantaneous feedback to satisfy the human perception of conversation, whereas VoD does not. This
immediate feedback requires minimal delay and jitter on the network.

HUMAN PERCEPTION OF QOS

In this section we present the human perception of the QoS measures presented previously. These are
the perceptions from a user’s standpoint, which differ from the perception of these QoS aspects from a
technology standpoint.

Human Perception of Delay

In conventional text and data networking, delay requirements are the least stringent. The response time
in these types of applications can increase from 2 to 5 seconds before becoming unacceptable
(Szuprowicz, 1995). Even at 5 seconds, the response time may still be considered tolerable.
58 Y. Chen et al. / QoS Requirements of Network Applications on the Internet

End-to-end audio latency refers to the latency experienced across a link. In other words, the time
between when the sound is sent on one end of a link, and when it is received on the other. The
relationship between End-to-end Audio Latency and the Human Ear is shown in Table 1. (Szuprowicz,
1995).
In interactive applications of real time sound transmission, as well as in virtual reality, the overall one-
way delay needs to be short in order to give the user an impression of real time responses. A maximum
value on the order of 0.1 to 0.5 seconds is required to accomplish this goal. (Fluckiger, 1995).
Based on subject tests, the International Telecommunication Unit (ITU) G.114 specification
recommends less than a 150 millisecond one-way end-to-end delay for high-quality real time traffic in
telecommunication. (ITU G.114, 1996) The G.114 time limits are shown in Table 2.

Table 1
End-to-end Audio Delay and the Human Ear

Audio Delay (ms) Effect of Delay on Human Voice Perception

> 600 Speech is unintelligible and incoherent
600 Speech is barely coherent
250 Speech is annoying but comprehensible
100 Imperceptible different between audio and real speech
50 Humans cannot distinguish between audio and real speech

Table 2
G.114 Limits for One-way Transmission Time

One-way transmission time Effect of Delay on Human Voice Perception

0 to 150 ms Acceptable for most users
150 to 400 ms Acceptable, but had impact
400 ms and above Unacceptable

In this paper, we adopt the G.114 limit of delay, 150 milliseconds, for most real time traffic that does
not require a high level of interaction, such as teleconferencing, and streaming audio and video. On the
contrary, interactive applications like Voiceover IP (VoIP) and Videophony require the value of delay
given in Table 1 of 100 milliseconds due to the desire for instantaneous feedback.
In video applications, it is necessary to preserve the timing relationships between audio and video
streams, as well as the timing relationships within individual video streams. First, we consider the way in
which human beings perceive sounds and images.
Humans are much more sensitive to alteration of audio than of visual signals (Fluckiger, 1995). Lip-
synchronization is defined as the difference between the visual reception of a video, and the auditory
reception of its sound. For adequate user perception, lip-synchronization should not exceed 100ms
(Fluckiger, 1995)
In many applications, an audio and a video stream are transmitted concurrently. Compressor/
Decompressor (codec) technology processes audio and video data. In some codec technology, audio and
video data are put into separate packets. In such cases, the audio stream must have priority over the video
stream. Since the delay of a real time stream should be less than 150 milliseconds, the delay of a video
stream should not exceed 250 milliseconds to preserve the 100 milliseconds limit on lip-synchronization.
In other codec technologies, audio and video data are put into the same packets. Under these
circumstances, we have to satisfy the delay requirement of the audio stream, which has a maximum limit
 Y. Chen et al. / QoS Requirements of Network Applications on the Internet 59

of 150 milliseconds. In this paper, we consider the case in which audio and video data are in the same
packets.

Human Perception of Jitter

Jitter is an essential performance parameter of a network intended to support real time sound and image
media. Of all information types, real time sound is the most sensitive to network jitter.
The solution to overcoming jitter in real time sound transmissions is for the receiving system to wait a
sufficient length of time, called a delay offset, before playing received sounds. Incoming blocks are stored
in a temporary memory location, called a buffer. After an adequate delay, these sound blocks can be
played smoothly. This process is called delay equalization or delay compensation.
One of the shortcomings of the delay equalization technique is that of the additional delay, which is
introduced at the sink end. Therefore, the value of jitter has a close relationship with the delay
requirement of a specific application.
With typical personal computers or workstations as end-systems, the variation of the network transit
delay should, in general, not exceed 100ms for CD-quality compressed sound and 400ms for telephone-
quality speech. For multimedia applications with strong bounds on transit delay, like virtual reality, this
jitter should not exceed 20–30 ms (Fluckiger, 1995).
Video applications have different human perception requirements for the minimization of jitter. For
example, the variation of network transit delay should not exceed 50ms for HDTV quality, 100ms for
broadcast quality, and 400ms for videoconferencing (Fluckiger, 1995).

Human Perception of Error

Three considerations must be taken into account when defining the acceptable human requirements
for error. The first is current retransmission techniques. In conventional text and data transmissions,
retransmission techniques are employed to correct errors in transmitted data. In the case of real time
audio and video, retransmission techniques are not appropriate. Secondly, we need to realize that
humans usually tolerate a high level of transmission error. This tolerance is generally derived from
human experiences in comparable situations. Finally, in the case of compressed bit streams, erroneous
or lost packets may invalidate greater numbers of subsequent packets. As a result, lower values of
network error and loss rates must be maintained when audio and video data are in a higher compression
rate scheme.
Furthermore, limits on error rates vary by application. The manner in which data are presented to
humans (data type) dictates limits on error rates. In the case where data are presented to human users
without recording for further processing, the residual bit error rate of a telephone-quality audio stream
should be lower than 10−2. The residual bit error rate of a CD-quality audio stream should be lower than
10−3 in the case of an uncompressed format, and lower than 10−4 in the case of a compressed format
(Fluckiger, 1995).
The end-to-end network bit error rate, before possible error recovery between end-systems, should not
exceed 10−6 for HDTV quality, 10−5 for broadcast TV quality, and 10−4 for videoconference quality
(Fluckiger, 1995). These figures are for compressed video data streams.

CLASSIFICATION OF APPLICATIONS

Based on the different technology attributes and human perception characteristics discussed in this
paper, we propose the taxonomy of traffics on network in this section. The technology attributes
60 Y. Chen et al. / QoS Requirements of Network Applications on the Internet

previously defined in this paper serve as the basis for classifying multimedia applications into four
categories based on time dependence and symmetry.

Non Real Time & Asymmetric

Non real time, asymmetric applications generate traffic that is considered “best effort” traffic. This
means that the traffic is transmitted using a best effort protocol. The best effort protocol does not provide
any specific QoS reliability and simply processes traffic on a first-come, first-served basis. The most
common NRT, asymmetric network applications are as follows:
Web Browsing — Hyper Text Transport Protocol (HTTP)
Enhanced Web Browsing
Email
File Transfer Protocol (FTP)
Telnet.

Non Real Time & Symmetric

Most common NRT network applications are asymmetric. This makes sense because NRT traffic
usually consists of client machines requesting data and services from a host machine. However, there is
one commonly used NRT application that is symmetric in nature and thus requires equivalent resource
consumption at each host endpoint of the communication link:
Internet Relay Chat.

Real Time & Asymmetric

Real time applications inherently have more stringent QoS requirements due to the nature of real time
transmissions. Common real time, asymmetric applications are as follows:
Audio Broadcasting
Video Broadcasting
Interactive Audio on Demand
Interactive Video on Demand
Telemetry.

Real Time & Symmetric

Common real time, symmetric applications include those which are conversational in nature:
Teleconferencing (including Audio, Audiographics, and Video Conferencing)
Videophony
VoIP.

QOS METRICS FOR EACH CLASS OF APPLICATION

In this section we systematically propose the desired QoS metrics for each application listed in the
previous section. We base these metrics on knowledge about human perception of QoS and the
technology features of various traffics put forth in this paper. Capacities of networking equipment are
described in Appendices I and II, which can be used to examine the capability of the current networking
technology in supporting QoS metrics.
 Y. Chen et al. / QoS Requirements of Network Applications on the Internet 61

QoS Metrics for Web Browsing

The QoS metrics for conventional web-browsing using HTTP are outlined in Table 3.

Table 3
QoS Metrics for Web Browsing

QoS Metrics
Timeliness Preciseness Accuracy
Traffic Technology Response
Class Attributes Required
time Delay Jitter Data Rate Loss
Bandwidth Error Rate
Expected (ms) (ms) (bps) Rate
(bps)
by Users
Web Non Real Time 2–5
< 400 N/A < 30.5 K < 30.5 K Zero Zero
Browsing and Asymmetric Seconds
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

Requirements for conventional web browsing are mainly influenced by response time, which is limited
to no more than 5 seconds. Delay less than 400 ms is expected for all best effort network traffic although,
under some circumstances, delays greater than 400 ms may be considered acceptable. Jitter is not
applicable in HTTP because it has little impact on traditional static text and picture web browsing. Data
rate and required bandwidth for typical applications of web browsing are less than 30.5 kbps. The
expected loss rate and error rate are zero since HTTP is a reliable transfer protocol in which erroneous
packets are sent again using a retransmission policy.

QoS Metrics for Enhanced Web Browsing

Enhanced web browsing refers to high-priority transaction services, such as those related to e-
commerce. The primary performance requirement for enhanced HTTP is to provide a sense of
immediacy to the user. This sense of immediacy acts as a security blanket by assuring the user that a
transaction is preceding smoothly. The QoS metrics for enhanced web browsing are outlined in
Table 4.

Table 4
QoS Metrics for Enhanced Web Browsing

QoS Metrics
Timeliness Preciseness Accuracy
Traffic Technology Response
Class Attributes Required
time Delay Jitter Data Rate Loss
Bandwidth Error Rate
Expected (ms) (ms) (bps) Rate
(bps)
by Users
Enhanced Non Real
2–4
Web Time and < 400 N/A < 24 K < 24 K Zero Zero
Seconds
Browsing Asymmetric
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II
62 Y. Chen et al. / QoS Requirements of Network Applications on the Internet

For enhanced web browsing, a value of 2–4 seconds for response time is suggested as acceptable
to most users. The best effort traffic delay specification of less than 400 ms is sufficient to satisfy
this 2–4 second response time requirement. Again, jitter is not applicable to enhanced HTTP.
Compared with traditional web browsing, enhanced web browsing requires less bandwidth. This is
due to the fact that web pages used in a transaction service are typically simpler and more practical
than web pages that are customarily browsed. However, this bandwidth requirement may be
increased by adding security measures, such as encryption and SSL. The expected loss rate and error
rate are zero.

QoS Metrics for Email

Email is a store and forward type of service which, in principle, can tolerate delays of several
minutes or even hours from a technology perspective. However, it is important to differentiate
between communications between the user and the local email server and server to server transfers.
When a user communicates with a local mail server, there is an expectation that the mail will be
transferred quite rapidly, although not necessarily instantaneously. The QoS metrics for email are
given in Table 5.

Table 5
QoS Metrics for Email

QoS Metrics
Timeliness Preciseness Accuracy
Traffic Technology Response
Class Attributes Required
time Delay Jitter Data Rate Loss
Bandwidth Error Rate
Expected (ms) (ms) (bps) Rate
(bps)
by Users
Non Real
2–5
Email Time and Low N/A < 10 K < 10 K Zero Zero
Seconds
Asymmetric
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

Consistent with our research findings on response time for web browsing, a requirement of 2–5
seconds for email response time is proposed. The requirement of expected network delay is low, and jitter
is not applicable. The typical bandwidth requirement for email is less than 10 kbps. The expected loss rate
and error rate are zero.

QoS Metrics for FTP

Typically, FTP service requires a relatively high bandwidth, which is clearly influenced by the size of
the file. Meanwhile, the size of files will also impact the time expected to finish the service. We also need
to differentiate between the time elapsed between command transmissions, and file transmissions between
the user and the FTP server. User expectation is that a file should begin transmission soon after the
command sent out. QoS metrics for FTP are given in Table 6.
Again, a requirement for response time of 2–5 seconds is proposed based on our web browsing
research findings. As for delay, as long as there is some indication that the file transfer is proceeding, it is
reasonable to assume a somewhat greater tolerance for delay than the expected delay for a single web
page. Jitter is not applicable, and the expected loss rate and error rate are zero.
 Y. Chen et al. / QoS Requirements of Network Applications on the Internet 63

Table 6
QoS Metrics for FTP

QoS Metrics
Timeliness Preciseness Accuracy
Traffic Technology Response
Class Attributes Required
time Delay Jitter Data Rate Loss
Bandwidth Error Rate
Expected (ms) (ms) (bps) Rate
(bps)
by Users
Non Real
2–5
FTP Time and Med N/A High High Zero Zero
Seconds
Asymmetric
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

QoS Metrics for Telnet

Requirements for quality of service of telnet applications are focused on a short delay in order to
provide essentially instantaneous echo-back of characters. The QoS metrics for telnet are given in
Table 7.

Table 7
QoS Metrics for Telnet

QoS Metrics
Timeliness Preciseness Accuracy
Traffic Technology Response
Class Attributes Required
time Delay Jitter Data Rate Loss
Bandwidth Error Rate
Expected (ms) (ms) (bps) Rate
(bps)
by Users
Non Real
<2
Telnet Time and < 250 N/A <1K <1K Zero Zero
seconds
Asymmetric
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

For telnet applications, a response time of less than 2 seconds is suggested in order to satisfy the human
perception of users and a delay less than 250 millisecond is recommended. Since most of the data
transmitted via telnet are plain characters, the requirement for bandwidth is quiet small, typically less than
1 kbps. Jitter is not applicable, and the expected loss and error rates are zero.

QoS Metrics for Internet Relay Chat

Internet Relay Chat is a multi-user chat system, where users convene on “channels” (a virtual place,
usually with a topic of conversation) to talk in groups, or privately. Requirements for quality of service in
Internet Relay Chat are focused on short response times in order to provide essentially instantaneous
character feedback. The QoS metrics for Internet Relay Chat are given in Table 8.
64 Y. Chen et al. / QoS Requirements of Network Applications on the Internet

Table 8
QoS Metrics for Internet Relay Chat

QoS Metrics
Timeliness Preciseness Accuracy
Traffic Technology Response
Class Attributes Required
time Delay Jitter Data Rate Loss
Bandwidth Error Rate
Expected (ms) (ms) (bps) Rate
(bps)
by Users
Non Real
Internet
Time and 1 Second < 200 N/A <1K <1K Zero Zero
Relay Chat
Symmetric
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

A response time of less than 1 second is suggested to satisfy users’ expectations. A delay less than 200
milliseconds is recommended to guarantee a short response time. In general, Internet Relay Chat consists
of plain character transmission, so the requirement for bandwidth is quiet small, typically less than 1
kbps. Jitter is not applicable in Internet Relay Chat. With the existence of a retransmission policy for
erroneous packets, the expected loss rate and error rate are zero.

QoS Metrics for Audio Broadcasting

User response time in Audio Broadcasting refers to the time elapsed between transmitting commands and
receiving feedback from the audio server. The QoS metrics for audio broadcasting are given in Table 9.

Table 9
QoS Metrics for Audio Broadcasting

QoS Metrics
Timeliness Preciseness Accuracy
Technology Response
Traffic Class Data Required
Attributes time Delay Jitter Loss
Rate Bandwidth Error Rate
Expected (ms) (ms) Rate
(bps) (bps)
by Users
Real Time
Audio 2–5
and Highly <150 <100 56–64K 60–80 K <0.1% <0.1%
Broadcasting Seconds
Asymmetric
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

For audio broadcasting we adopt a one-way end-to-end delay for high-quality real time traffic from
ITU G.114 specifications of less than 150 milliseconds. For jitter, no more than 400 ms is recommended
for telephone-quality speech and no more than 100 ms is recommended for CD-quality compressed sound
(Fluckiger, 1995). We suggest a jitter limit of no more than 100 ms for broadcast quality sound.
A typical data stream for audio broadcasting is 56–64 kbps (Kenyon and Nightingale, 1992). The
corresponding bandwidth requirement is around 60–80 kbps. The residual bit error rate is recommended
to be lower than 10−2 for a telephone-quality audio stream and lower than 10−4 for a CD-quality audio
stream using a compressed format (Fluckiger, 1995). Since there is a higher quality requirement in
 Y. Chen et al. / QoS Requirements of Network Applications on the Internet 65

broadcasting than in telephone, and a lower quality requirement than in a CD audio stream, we
recommend a loss rate and error rate of no more than 10−3 for broadcast audio streams.

QoS Metrics for Video Broadcasting

In Video Broadcasting, under different codec technologies and quality requirements, there are different
requirements for network transmissions. We list the QoS metrics for video broadcasting using some
typical codec technologies in Table 10.

Table 10
QoS Metrics for Video Broadcasting

QoS Metrics
Timeliness Preciseness Accuracy
Technology Response
Traffic Class Required
Attributes time Delay Jitter
n/a Bandwidth Loss Rate Error Rate
Expected (ms) (ms)
(bps)
by Users
2–5
Real Time and Seconds
Video
Highly <150 <100
Broadcasting
Asymmetric Lip-synch:
<100ms
Typical Application Coding Standard
VCR Quality MPEG-1 <100 1.2–1.5M <0.001% <0.001%
Video Quality slightly superior
to that of broadcast TV (NTSC 4–60 M
or PAL) with bit rate of 4M MPEG-2
HDTV requiring bit rate from
<50 <0.0001% <0.0001%
15–34 M
Multimedia on Web MPEG-4 <150 28.8–500K <0.001% <0.001%
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

For VCR quality MPEG-1 video streams, the required bandwidth is around 1.2–1.5 Mbps (Silveira
et al, 1996). Jitter is recommended to be less than 100 ms for a broadcast-quality video stream
(Fluckiger, 1995). The residual bit error rate is recommended to be lower than 10−5 for a broadcast
TV quality video stream using a compressed format (Fluckiger, 1995). The loss and error rates
should be less than 10−5.
For HDTV quality MPEG-2 video streams, the required bandwidth is around 40 Mbps (Silveira et al,
1996). Jitter is recommended to be less than 50 ms for HDTV quality video streams (Fluckiger, 1995).
The residual bit error rate is recommended to be lower than 10−6 for HDTV quality video streams using
a compressed format (Fluckiger, 1995). The loss rate and error rate should be less than 10−6.
For typical MPEG-4 video streams on the World Wide Web, the required bandwidth is around
28.8–500 kbps (Silveira et al, 1996). Jitter less than 150 ms is recommended due to a lower quality
requirement than VCR technologies, which require a jitter less than 100ms. On the other hand,
MPEG-4 also has a higher compression rate, which usually requires less residual error. Based on
considerations of quality and compression rate, we recommend a loss rate and error rate less than
10−5.
66 Y. Chen et al. / QoS Requirements of Network Applications on the Internet

QoS Metrics for Interactive Audio on Demand

Interactive audio-on-demand on the World Wide Web is the ability to listen to sound on a personal
computer (with a sound card and speakers installed) while the file is passed through an individual modem.
Under different codec technologies and quality requirements, there are different requirements for network
transmission. We list the QoS metrics for some typical audio-on-demand applications in Table 11.

Table 11
QoS Metrics for Interactive Audio on Demand

QoS Metrics
Timeliness Preciseness Accuracy
Technology Response
Traffic Class Required
Attributes time Delay Jitter Loss
n/a Bandwidth Error Rate
Expected (ms) (ms) Rate
(bps)
by Users
Interactive Real Time and
2–5
Audio on Highly <150 <100
Seconds
Demand Asymmetric
Typical Application Coding Standard
MP3, 64K for mono,
MPEG-1 32–448K <0.1% <0.1%
128K for stereo
Advanced Audio Coding
(AAC) providing CD quality MPEG-2 384K <0.01% <0.01%
audio
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

For a mono-quality MP3 audio stream, the required bandwidth is 64 kbps. For a stereo-quality MP3
audio stream, the required bandwidth is 128 kbps. Delay less than 150ms is recommended for high-
quality real time traffic in the ITU G.114 standard (ITU, 1996). A jitter rate of less than 100 ms for CD-
quality compressed sound is recommended (Fluckiger, 1995). The residual bit error rate is recommended
to be lower than 10−3 for a broadcasting-quality audio stream in a compressed format. Therefore, the loss
rate and error rate should be less than 10−3.
For a CD-quality MP2 audio stream, the required bandwidth is 384 kbps. A delay less than 150ms is
recommended for high-quality real time traffic in the ITU G.114 standard (ITU, 1992). A jitter rate of less
than 100 ms for CD-quality compressed sound is recommended (Fluckiger, 1995). The residual bit error
rate is recommended to be lower than 10−4 for a CD-quality audio stream in a compressed format.
Therefore, the loss rate and error rate should be less than 10−4.

QoS Metrics for Interactive Video on Demand

Interactive video-on-demand refers to a range of applications whereby users can request access to
video servers, which provide still and moving pictures on an individual basis (Fluckiger, 1995). Elapsed
time in video-on-demand refers to the time elapsed from a request generation to the start of video
playback. The QoS metrics for video-on-demand are given in Table 12.
Elapsed time of a few minutes should be acceptable for video-on-demand because users are not
particularly concerned with the starting delay, but instead are concerned with the delay during playback.
Response to interactive functions, such as play and next, should be less than 2–5 seconds. The typical
 Y. Chen et al. / QoS Requirements of Network Applications on the Internet 67

bandwidth requirement for the user’s local network is 1.5 Mbps. The bandwidth requirement for the
backbone depends on how many users the video server needs to support simultaneously. In general,
backbone bandwidth is hundreds of megabits per second.
Other QoS metrics for video-on-demand are the same as those for video broadcasting.

Table 12
QoS Metrics for Interactive Video on Demand

QoS Metrics
Timeliness Preciseness Accuracy
Traffic Technology
Class Attributes Response Required
Delay Jitter Loss Error
time Expected n/a Bandwidth
(ms) (ms) Rate Rate
by Users (bps)
Elapsed
Time: a few
minutes Local:1.5M
(typical
Interactive Real Time Response to application)
Video on and Highly Interactive <150
Demand Asymmetric function: Backbone:
2–5 Seconds hundreds of
mega
Lip-synch:
<100ms
Typical Application Coding Standard
1.2–1.5M / A
VCR Quality MPEG-1 <100 <0.001% <0.001%
Single Video
Video Quality slightly
superior to that of 4–60 M / A
broadcast TV (NTSC or Single Video
MPEG-2
PAL) with bit rate of 4M
HDTV requiring bit rate <0.0001 <0.0001
<50
from 15–34 M % %
28.8–500K / A
Multimedia on Web MPEG-4 <150 <0.001% <0.001%
Single Video
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

QoS Metrics for Telemetry

Telemetry refers to the telemetering and telecontrol of industrial processes. Telemetry is an example of
a data service which requires real time streaming performance. Clearly, two-way control implies a tight
limit on the allowable delay in telemetry. The QoS metrics for telemetry are given in Table 13.
For telemetry, a delay value of 250 milliseconds is proposed (QoS Performance Requirement for
UMTS (Nortel, 2001)). The range of bandwidth requirements is between 2 kbps and 52 Mbps, depending
on the specific telemetry system. A key differentiator from voice and video services in this category is the
zero tolerance for information error and loss (QoS Performance Requirement for UMTS (Nortel, 2001)).
It is clearly understandable that there cannot be data error and loss when controlling an important
industrial process.
68 Y. Chen et al. / QoS Requirements of Network Applications on the Internet

Table 13
QoS Metrics for Telemetry

QoS Metrics
Timeliness Preciseness Accuracy
Traffic Technology
Class Attributes Response time Required
Delay Jitter Loss
Expected by n/a Bandwidth Error Rate
(ms) (ms) Rate
Users (bps)
Real Time and
Telemetry Highly <250 <100 2K–52M Zero Zero
Asymmetric
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

QoS Metrics for Teleconferencing

Teleconferencing can occur in three forms: audio conferencing, audiographics conferencing, and video
conferencing. Each of these forms requires its own QoS metrics. The metrics for the three forms of
teleconferencing are outlined in this section.

QoS Metrics for Audio Conferencing

Audio Conferencing provides an audio link similar to that of a conventional telephone, except that it
offers much higher quality audio and enables more than two sites to be linked together. Using hands-free
audio units with sensitive microphones and sophisticated echo-cancellation software, audio-conferencing
enables communication between groups of participants rather than just individuals (Bleazard, 1985).
Voice data transmission dictates the QoS metrics of audio conferencing shown in Table 14.

Table 14
QoS Metrics for Audio Conferencing

QoS Metrics
Timeliness Preciseness Accuracy
Technology Response
Traffic Class Required
Attributes time Delay Jitter Data Rate Loss
Bandwidth Error Rate
Expected (ms) (ms) (bps) Rate
(bps)
by Users
Audio Real Time and
<150 <400 <1% <1%
Conferencing Symmetric
Coding Standard
G.711 64 K 80 K
G.726 40~16 K 50–22 K
G.728 16 K 22 K
G.729 8K 11 K
G.723.1 6.3/5.3 K 9/8K
GSM FR 13 K 18 K
GSM EFR 12.2 K 17 K
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II
 Y. Chen et al. / QoS Requirements of Network Applications on the Internet 69

ITU G.114 specification recommends less than a 150 millisecond one-way end-to-end delay for high-
quality real time traffic, such as voice transmissions. Jitter of no more than 400 milliseconds is
recommended for telephone-quality speech (Fluckiger, 1995). In the case of real time audio and video,
retransmission techniques are not appropriate. G.729 codec requires packet loss far less than 1 percent to
avoid audible errors. In the case of presentation to human users without recording for further processing,
the residual bit rate error rate of a telephone-quality audio stream should be lower than 10−2 (Fluckiger,
1995). Differing Codec technologies lead to different bandwidth requirements.

QoS Metrics for Audiographics Conferencing

Except for the audio link, audiographics conferencing enables participants at two or more sites to have
a shared workspace on their computer desktops. This might be a shared “whiteboard” where they can
draw, write, or import and manipulate images collaboratively in real time. Audiographics conferencing
might also be “application sharing”, where a piece of software can be run and controlled by both users (G
B Bleazard, 1985). This type of conferencing is useful when users at different sites want to work together
on documents such as reports or statistical data in spreadsheets. The largest QoS concern for
audiographics is voice quality. The QoS metrics for audiographics conferencing are given in Table 15.

Table 15
QoS Metrics for Audiographics Conferencing

QoS Metrics
Timeliness Preciseness Accuracy
Technology
Traffic Class Required
Attributes Delay Jitter Loss
n/a n/a Bandwidth Error Rate
(ms) (ms) Rate
(bps)
Audiographics Real Time and
<150 <400 9.6–19.6K <1% <1%
Conferencing Symmetric
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

ITU G.114 specification recommends less than a 150 millisecond one-way end-to-end delay for high-
quality real time traffic, such as voice. For jitter, a rate of no more than 400 milliseconds is recommended
for telephone-quality speech (Fluckiger, 1995). The G.729 codec requires a packet loss rate of far less
than 1 percent to avoid audible errors. In the case of presentation to human users without recording for
further processing, the residual bit error rate of a telephone-quality audio stream should be lower than
10−2 (Fluckiger, 1995). Typical bandwidth requirements for audiographics conferencing are 9.6–19.6
kbps (Bleazard, 1985).

QoS Metrics for Video Conferencing

Video conferencing enables real time communication over a distance by allowing people at two or
more sites to communicate with each other. In video conferencing, in addition to hearing each others
voices, as with a conventional telephone, people can see video pictures of each other from their various
sites (Janssen et al, 2001). Each site has one or more cameras, microphones, loudspeakers and monitors,
as well as a codec. This type of conferencing aims to create a sense of people from distant sites being in
the same room, an effect that has been called Virtual Presence. Video conferencing is not only concerned
with voice quality, but is also concerned with video quality. The QoS metrics for video conferencing are
given in Table 16.
70 Y. Chen et al. / QoS Requirements of Network Applications on the Internet

Table 16
QoS Metrics for Video Conferencing

QoS Metrics
Timeliness Preciseness Accuracy
Technology Response
Traffic Class Required
Attributes time Delay Jitter Data Rate Loss Error
Bandwidth
Expected (ms) (ms) (bps) Rate Rate
(bps)
by Users
Video Real Time and Lip-synch:
<150 <400 <0.01% < 0.01%
Conferencing Symmetric <100 ms
Coding Standard
H.320 64–1920K 80K–2M
H.323 64X K 80X K
H.324 <64K <80K
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

Since humans are much more sensitive to alteration of audio signals than visual signals, we adopt the
recommendations for delay and jitter from ITU G.114: less than a 150 millisecond one-way end-to-end
delay for high-quality real time traffic and no more than a 400 millisecond jitter for telephone-quality
speech (Fluckiger, 1995). However, the requirement for loss rate and error rate differ. Recall that
compressed streams are more sensitive to errors than uncompressed streams since one lost packet may
influence a number of subsequent packets. Similarly, streams with a higher compression ratio are more
sensitive to errors than streams with a lower compression ratio since more packets will be impacted.
Therefore, the end-to-end network bit error rate for compressed streams, before possible error recovery,
between end-systems is recommended not to exceed 10−4 for videoconference quality (Fluckiger,
1995).
The H.320-series governs the basic video-telephony concepts of audio, video and graphical
communications by specifying requirements for processing audio and video information, providing
common formats for compatible audio/video inputs and outputs, and defining protocols that allow a
multimedia terminal to utilize the communication links and synchronization of audio and video signals.
H.320 standards address ISDN Videoconferencing. The H.323 standards address Video communications
on Local Area Networks, and the H.324 standards address video and audio communications over low bit
rate connections such as POTS modem connections. From Table 16, we can see that different technology
standards lead to different bandwidth requirements.
It is worthy to note that the bandwidth requirements for video conferencing and videophony (described
in the next section) may be less stringent than that of general video. This is based on the assumption that
these media will likely have less motion and little or no scene changes in the picture.

QoS Metrics for Videophony

Videophony is a companion audiovisual service of videoconferencing. It may be distinguished by the

following features (Kenyon and Nightingale, 1992): it is primarily for person-to-person rather than group-
to-group audiovisual communication, and it is usually an on-demand service provided on customer-
switched networks, whereas videoconferences can be scheduled in advance. Compared with video
conferencing, videophony is a more interactive application, which requires instantaneous responses from
dialogists. The QoS metrics for videophony are given in Table 17.
 Y. Chen et al. / QoS Requirements of Network Applications on the Internet 71

Videophony requires a shorter end-to-end delay of 100 ms to make the difference between audio and real
speech imperceptible to users (Fluckiger, 1995). Beyond 100 ms, although speech is still comprehensible,
users may become irritated with the service. Since videophony is a special application of video conferencing, it
complies with the same technology standards and has the same QoS metrics as video conferencing.

Table 17
QoS Metrics for Videophony

QoS Metrics
Timeliness Preciseness Accuracy
Traffic Technology Response
Class Attributes Required
time Delay Jitter Data Rate Loss Error
Bandwidth
Expected by (ms) (ms) (bps) Rate Rate
(bps)
Users
Phone to
Phone Delay:
Real Time
<150 ms
Videophony and <100 <400 <0.01% < 0.01%
Symmetric
Lip-synch:
<100 ms
Coding Standard
H.320 64–1920K 80K–2M
H.323 64X K 80X K
H.324 <64K <80K
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II

Table 18
QoS Metrics for Voiceover IP

QoS Metrics
Timeliness Preciseness Accuracy
Traffic Technology
Class Attributes Response time Required
Delay Jitter Data Rate Loss Error
Expected by Bandwidth
(ms) (ms) (bps) Rate Rate
Users (bps)
Real Time Phone to Phone
Voiceover
and Delay: <100 <400 < 1% < 1%
IP
Symmetric <150 ms
Coding Standard
G.711 64 K 80 K
G.726 40~16 K 50–22 K
G.728 16 K 22 K
G.729 8K 11 K
G.723.1 6.3/5.3 K 9/8K
GSM FR 13 K 18 K
GSM EFR 12.2 K 17 K
Network Capacity
Link: Refer to Appendix I Router: Refer to Appendix II
72 Y. Chen et al. / QoS Requirements of Network Applications on the Internet

QoS Metrics for Voiceover IP

The QoS metrics for Voiceover IP are given in Table 18.

Like videophony, VoIP also requires a shorter end-to-end delay of 100 ms to give the users an
imperceptible difference between audio and real speech (Szuprowicz, 1995). A jitter rate of less than 400
milliseconds is suggested for telephone-quality speech (Fluckiger, 1995). We adopt the loss rate
requirement, packet loss far less than 1 percent, in the G.729 codec. In the case of presentation to human
users without recording for further processing, the residual bit rate error rate of a telephone-quality audio
stream should be lower than 10−2 (Fluckiger, 1995). Again, different voice coding technologies lead to
different bandwidth requirements.

SUMMARY

In order for users to find a certain type of multimedia service acceptable, the application must meet
specific QoS requirements. Metrics used to determine QoS include response time, delay, jitter, data rate,
required bandwidth, loss rate and error rate. In order to provide QoS on a network, requirements for QoS
must be established. Defining the limits of QoS metrics helps to establish necessary requirements.
In this paper, we consider two key factors of QoS requirements: human factor and technology
attributes. Based on our survey of human factors and technology features, we propose numeric QoS
metrics of network applications in this paper. Up to date, we have not found any literature stating metrics
which differed from our findings. Based on the lack of literature which would dispute these findings, we
could not definitively state whether or not specific measures are “widely accepted”. However, we believe
that metrics proposed in this paper provide a starting point to establish a universally accepted set of QoS
requirements for network applications on the Internet. We concede that these proposed measures could be
refined through experimental verification. Further work can make use of these metrics to differentiate
network traffic for realizing quality assurance in networks.

ACKNOWLEDGEMENTS

This work is sponsored by the Air Force Research Laboratory – Rome (AFRL-Rome) under grant
number F30602-01-1-0510, and the Department of Defense (DoD) and the Air Force Office of Scientific
Research (AFOSR) under grant number F49620-01-1-0317. The U.S. government is authorized to
reproduce and distribute reprints for governmental purposes notwithstanding any copyright annotation
thereon. The views and conclusions contained herein are those of the authors and should not be
interpreted as necessarily representing the official policies or endorsements, either express or implied, of,
AFRL-Rome, DoD, AFOSR, or the U.S. Government.

REFERENCES
Austerberry, D. (2002). The Technology of Video & Audio Streaming. Focal Press.
Bleazard, G.B. (1985). Introducing Teleconferencing. NCC Publication. pp. 58.
Eslinger, B. & Baker, M. (2002). Advances in Telemetry and Network Transmission, International Test and
Evaluation Association (ITEA). Instrumentation Conference.
ETSI TIPHON (2000). End-to-end Quality of Service in TIPHON Systems. Part 2: Definition of Quality of Service
(QoS) Classes. ETS. 101 329–2.
 Y. Chen et al. / QoS Requirements of Network Applications on the Internet 73

Fiedler, U. & Plattner, B. Quality of Service in Business-to-Business E-Commerce Application. Computer

Engineering and Networks Laboratory, Switzerland
Fluckiger, F. (1995). Understanding Networked multimedia. Prentice Hall, pP. 242–382.
Janssen, J.,Vleeschauwer, D.D., Buchli, M.J.C. & Kooij, R.E. (2001). Tuning an IP-based Network Transporting
Telephony and Videophony. Alcatel Bell, Network Strategy Group, Francis Wellesplein 1, B-2018 Antwerp,
Belgium.
Kenyon, N.D. & Nightingale, C. (1992). Audiovisual Telecommunications. BT Laboratories, Ipswich, UK, Chapman
& Hall.
Ko, M. & Koo, I. (1996). An Overview of Interactive Video On Demand System. The University of British
Columbia.
Menasce, D.A. & Almeida, V.A.F. (1998). Capacity Planning for Web Performance. Prentice Hall, PTR.
Nortel Networks (2001). QoS Performance Requirement for UMTS. Retrieved November 8, 2001:
http://www.3gpp.org/ftp/tsg_sa/WG1_Serv/TSGS1_03HCourt/Docs/Docs-Long_Names/
S199362_QoS_Nortel_Performance_requirements_for_UMTS.doc
Silveira, R.M., Margi, C.B., Gonzalez, L.G., Favero, E., Vilcachagua, O.D., Bressan, G. & Ruggiero, W.V.A.
(1996). Multimedia on Demand System for Distance Education. LARC-Laboratory of Computer Architecture
and Networks, PCS-Department of Computer and Digital System Engineering EPUSP-Polytechnic School,
University of San Paulo, Standards: ITU-T G.114.
Szuprowicz, B.O. (1995). Multimedia Networking. McGraw-Hill, Inc., 161–162.
To, T.P.J. & Hamidzadeh, B. (1998). Interactive Video-on-demand Systems. Kluwer Academic Publishers, pp. 9–15.
Marziale, V. & Vitaletti, A. (2001). A Framework for Internet QoS Requirements. IST Mobile Communication
Summit 2001, Barcelona, Spain.
Ye, N. (2002). QoS-centric stateful resource management in information systems. Information Systems Frontiers.
4(2) 149–160.

Yan Chen is a doctoral student in the Department of Industrial Engineering at Arizona State
University. She received a B.E. degree and a M.E. degree from Xiamen University, Xiamen,
Fujian, P.R.China, both in Systems Engineering, where her interests were primarily in the
application of Artificial Neural Network. Her current interests are in Modeling and Analysis of
Semiconductor Manufacturing and Scheduling of complex job shops.

Toni Farley is the Assistant Director of the Information and Systems Assurance Laboratory, and
a PhD student of Computer Science at Arizona State University (ASU), Tempe, Arizona. She
has been awarded a graduate fellowship from the AT&T Labs Fellowship Program (ALFP). She
received her BS degree in Computer Science from ASU in 2003. Her research interests include
computer networks, network security, and information assurance. She is a member of IEEE and
IEEE Computer Society.
74 Y. Chen et al. / QoS Requirements of Network Applications on the Internet

Nong Ye is a Professor of Industrial Engineering and an Affiliated Professor of Computer

Science and Engineering at Arizona State University. She holds a Ph.D. degree in Industrial
Engineering from Purdue University, a M.S. degree in Computer Science from the Chinese
Academy of Sciences, and a B.S. degree in Computer Science from Peking University. Her
research interests are in assuring process quality and preventing faults and errors in
information systems. Dr. Ye is an Associate Editor for IEEE Transactions on Reliability and
IEEE Transactions on Systems, Man, and Cybernetics. She is a senior member of the Institute
of Industrial Engineers and a senior member of IEEE.
 Y. Chen et al. / QoS Requirements of Network Applications on the Internet 75

Appendix I: Link Capacity of Networks

Wide Area Networks (WANs)

Technology used to build WANs Speed

X.25
ISDN
Frame Relay 45 M bps–several G bps
SMDS (Switched Multi-megabit Data Service)
ATM

Local Network (LANs)

Technology used to build LANs Speed Topology Medium Access Mechanism

10 M bps CSMA/CD (Carrier
Sense with Multiple
Ethernet Bus Twisted Pair
100 M bps Access / Collision
Detection)
4 M bps Shielded
Token Ring Ring Token-passing
16 M bps Twisted Cable
FDDI
Ring
(Fiber Distributed Data 100 M bps Optical Fiber Token-passing
(Two rings)
Interconnect)

LAN to WAN

Connection Speed
Dedicated leased line at T1 1.544 M bps
Dedicated leased line at T3 45 M bps

Home to WAN

Connection Speed
Dial-up Analog Modem 14.4 ~ 56 K bps
ISDN Basic Rate Interface (BRI) 128 K bps
ISDN Primary Rate Interface (PRI) 1.544 M bps
High-Bit-Rate Digital Subscriber Line (HDSL) 1.544 M bps
Asymmetric Digital Subscriber Line (ADSL) Outgoing: 640 K bps; incoming: 6 M bps
Cable TV Companies 10 M bps
 76 Y. Chen et al. / QoS Requirements of Network Applications on the Internet

Appendix II: — Router Capacity of Network

Router Upper Bandwidth

Cisco 12000 Series Internet Router 10 G bps
Cisco 10700 Series Internet Routers 5 G bps
Cisco 10000 Series Internet router 51.2 G bps
Cisco 7600 Series Internet Router 256 G bps
Cisco 7500 Series Internet Router 2 1.067 G bps
Cisco 7300 Series Internet Router 16 G bps
Cisco 7200 Series Internet Router 1.2 G bps
Cisco 7100 Series Internet Router 140 M bps

View publication stats