Xinping Guo Colin Pattinson
Experiences in the use of the Internet as a delivery medium for multimedia-based applications have revealed serious deficiencies in the ability of the communications medium to provide a data transmission service of sufficient quality, with consequential impact on the overall service quality experienced by users of the applications. This paper discusses the particular characteristics of multimedia traffic which make service provision so difficult; examines current proposals to develop a relationship between user perceptions of 'quality' and the factors which must be addressed by the service provider to deliver an acceptable level of service, and suggests the integration of modelling techniques to develop a more complete picture of the quality of service requirements.
It is a fundamental truth that the quality and quantity of any data delivered to a user is limited by the quality and quantity which the underling data transfer system(s) can support. It is also true that, in general, there is a mismatch between the theoretical maxima and those which are available in practice, due to factors such as the sharing of resources with other users, the overheads generated by the various communications tasks etc. The implications of this from a user's perspective are that any applications which rely on the transfer of data are limited by the ability of the data transfer systems in terms of speed, reliability and accuracy.
This places a requirement on the developers of these applications
to have an awareness of the impact of these limits, and to design
their systems accordingly. However, the situation is made more
complex by the fact that the quality of service(QoS) available varies
from network to network, and may also vary over time on the same
system, due to the need to share resources between a variable number
of other users. It is therefore important to be aware of the
processes by which QoS can be determined, negotiated and varied
before or during the operation of an application. This paper will
discuss the Quality of Service requirements for multimedia
communication systems, the use of source characterisation in resource
allocation, the characteristics of multimedia traffic, and QoS
requirements for multimedia traffic and its classification. Finally,
co-ordination between application and network based on QoS is
reviewed and general conclusions are drawn.
The traffic parameters describe traffic characteristics of the packetised data (or cell) streams . According to the CCITT, the following parameters are important in source characterisation[1]:
The above traffic characterisation parameters are used in
important network functions such as admission control, usage
parameter control and resource allocation. The values of the traffic
parameters are negotiated between the user/terminal and the network
during the call set-up phase: combined with the traffic
characteristics of the aggregate cell arrival stream in the network,
they are used for the operation of the admission control function,
deciding whether or not a new connection is to be accepted. In the
usage parameter control, the algorithm monitors whether the traffic
characterisation parameters negotiated during the connection
establishment are violated by the user during the call. Moreover, for
the resource allocation purposes, the traffic parameters are used by
the network operator as the basis for allocating resources to user
demands.
Source characterisation necessitates the precise definition of the behaviour of each source, and provides network management with the ability to manipulate flexibly the various services in terms of connection acceptance, negotiation of the QoS, congestion control, traffic enforcement and resource allocation.
The feasibility and efficiency of the QoS management architecture are strongly dependent upon the nature of the traffic to be accommodated. Selecting appropriate models of multimedia traffic sources is an important issue because traffic source characterisation provides network management with the ability to manipulate flexibly the various services in terms of connection acceptation, negotiation of the QoS, congestion control, traffic enforcement and resource allocation[1]. The multimedia traffic models range from constant bit-rate(CBR) to variable bit-rate(VBR). In particular, the variable bit rate (VBR) traffic generated by compressed audio and/or video is not only delay-sensitive but also bursty and long-range dependent[2]. Markov modulated process models[3], Fractional Brownian motion model[4] and Bounding Interval-Dependent(BIND) model[5] etc. have tried to capture the statistical characteristics of the aggregate and heterogeneous multimedia traffic. Further investigation and comparison of these traffic models is important. The target is to set up a better deterministic traffic model associated with a tighter traffic constraint function which in turn results in higher network utilisation.
Our studies [6] have shown that multimedia traffic differs from 'traditional' network traffic in the following ways :
These properties impose challenging QoS requirements for
multimedia communication systems; furthermore, different multimedia
applications present different demands on the network. Section 4 will
examine these requirements.
Recent experience with the Internet indicates that it is not well suited to time and loss sensitive multimedia applications such as voice and video. To support multimedia applications, the following six network criteria are critical :
These network criteria are closely associated with quality of service(QoS). According to ITU-T Recommendation E.800[7], the QoS is defined as follows:
The QoS is the collective effect of service performances which determine the degree of satisfaction of a user of the service.
This implies that the user is the final arbiter of 'good' or 'bad' QoS. Different applications demand different service qualities. Some need minimal delay and reliable response time, while others may need a good image quality. Table 1 summarises the five categories of QoS parameters[8]. The QoS is a difficult issue in that the relationship between application QoS parameters and network QoS parameters is very complex; QoS must be end-to-end; and the application QoS might change during connections. The following two sections explain such issues in detail.
|
Category |
Example Parameters |
|
Performance-oriented |
end-to-end delay and bit rate |
|
Format-oriented |
video resolution, frame rate, storage format, and compression scheme |
|
Synchronisation-oriented |
skew between the beginning of audio and video sequences |
|
Cost-oriented |
connection and data transmission charges and copyright fees |
|
User-oriented |
subjective image and sound quality |
QoS requirements for applications are typically end-to-end requirements which impose corresponding performance demands on both the network and the end-systems/applications. QoS parameters specify the resource quantity allocated to the service, as well as the service disciplines managing the shared resource. Therefore it is crucial to translate the user/application QoS into network QoS. The translation between user QoS and application QoS is nontrivial and still an open issue, because the perceptual issues are not completely understood[9]. Application QoS should be translated to the network QoS. The user specifies the application QoS. Then the communication system will map requirements into a set of system, protocol and network QoS specifications. This translation process is illustrated in Fig.1.
Fig.1 clearly shows that a user or an application specifies requirements; a communication system is responsible for meeting these specifications, and possibly requests an appropriate network resource to the network. The relationship between the application QoS and the network QoS is important because the application QoS can be very different from the network QoS. These QoS differences should be considered in the set-up stage between the user and the network provider. QoS parameters must be mapped to the resource requirements and the required resources must be determined, reserved and allocated along the path between the application and the provider/peer application.
Fig.1. QoS Mapping Diagram
To provide applications with end-to-end QoS guarantees, a QoS co-ordination architecture is required in which the QoS specification and QoS mechanism systematically provide application-level QoS requirement and end-to-end QoS guarantee[10]. From this point of view, the multimedia communication system should provide the following features:
In principle, the QoS co-ordination should configure, predict and maintain QoS specification to meet the end-to-end QoS. To manage QoS successfully, QoS co-ordination must be able to do the following [11] :
Allow explicit specification of QoS parameters when creating a session for multimedia transmission;
Current network QoS allocation systems are unable to answer these demands because they are incomplete, lack mechanisms to guarantee QoS and do not adhere to an overall framework [10].
No significant work has been reported so far on the integration of
multimedia traffic source modelling and network management with
emphasis on the quality of service guarantee[12][13]. Moreover, the
area of network support for flows remains an unsolved important issue
in the QoS management architecture. Therefore, it will be necessary
to design and develop an integrated QoS strategy, spanning traffic
source modelling and measurement, network scheduling disciplines and
resource management mechanisms. On-line estimation of data flow will
be applied to achieve accurate characterisation of the active
application. This estimation will then be employed to trigger dynamic
system and network resource reservation procedures. The current
network traffic load and the offered QoS will be reported to the user
to allow the choice or adjustment of QoS parameters while the
application is running. We intend that the results of our study will
promote the support of QoS provision in multimedia communication
systems.
Quality of service is increasingly important for multimedia
communication systems. This paper has addressed the important issues
of QoS provision for multimedia communication systems : first,
understanding multimedia traffic characterisation is very important
to identify the QoS requirement and implement QoS translation;
second, QoS requirement is indispensable for multimedia communication
systems; third, QoS translation involves the allocation of system and
network resources and this translation should be bi-directional for
the purposes of QoS negotiation and renegotiation; finally, QoS
co-ordination architecture must be required to provide guaranteed QoS
for the multimedia application during the connections. We envisage
that these functionalities of QoS will be truly supported in
multimedia communication systems (e.g. WWW's http protocol) in the
near future.
The authors are very grateful to Dr.Dave Hobbs and Dr.Becky
Strachan for their valuable discussions and comments.
[1] G.D.Stamoulis, M.E.Anagnostou and A.D.Georgantas, "Traffic Source Models for ATM Networks: a Survey", Computer Communications, Vol.17, No.6, June 1994, pp.428-438
[2] N.G.Duffield, J.T.Lewis and N.O'Connell, "Predicting Quality of Service for Traffic with Long-Range Fluctuations", Proceeding of IEEE ICC'95, Vol.1/3, 1995, pp.473-477
[3] U.R.Krieger, B.Muller-Clostermann and M.Sczittnick, "Modelling and Analysis of Communication Systems Based on Computational Methods for Markov Chains", IEEE J-SAC, Vol.8, No.9, Dec.1990, pp.1630-1648
[4] W.C.Lau, A.Erramilli, J.L.Wang and Walter Willinger, "Self-Similar Traffic Generation: The Random Midpoint Displacement Algorithm and its Properties", Proceeding of IEEE ICC'95, Vol.1/3, 1995, pp.466-477
[5] E.W.Knightly, "H-BIND: A New Approach to Providing Statistical Performance Guarantees to VBR Traffic", Proceedings of IEEE INFOCOM'96, Vol.3/3, 1996, pp.1091-1099
[6] X.P.Guo and C.Pattinson, "Comparison of Self-Similar Traffic Models", accepted by Fourth Communication Networks Symposium, Manchester, UK, 7-8 July, 1997
[7] Jae-ll Jung, "Quality of Service in Telecommunications Part I: Proposition of a QoS Framework and its Application to B-ISDN", IEEE Comm. Mag. Aug. 1996, pp.108-111
[8] Andreas Vogel, Brigitte Kerherve, Gregor von Bochmann and Jan Gecsei, "Distributed Multimedia and QOS: a Survey", IEEE Multimedia, Vol.2, No.2, Summer 1995, PP.10-19
[9] Jae-ll Jung, "Quality of Service in Telecommunications Part II: Translation of QoS Parameters into ATM Performance Parameters in B-ISDN", IEEE Comm. Mag. Aug. 1996, pp.112-117
[10]D.Hutchison, G.Coulson, A.Campbell and G.S.Blair, "Quality of Service Management in Distributed Systems", chapter 11 of the book "Network and Distributed Systems Management" (Edited by Morris Sloman), Addison-Wesley, 1995, pp.273-302
[11] I.Chung and T.Nakajima, "QoS Management in a Continuous Media Network System", 11th Conference Proceedings, Japan Society for Software Science and Technology, Oct. 1994
[12] Cristina Aurrecoechea, Andrew T.Campbell and Linda Hauw, "A Survey of QoS Architectures", Multimedia Systems Journal, Special Issue on QoS Architecture, 1997 (to appear)
[13] R.Nagarajan, "Quality of Service Issue in High Speed Networks", Ph.D.dissertation, University of Massachusetts, Sept. 1993