CASOPIS REPUBLICKE AGENCIJE ZA TELEKOMUNIKACIJE

The importance of Quality of Service (QoS) provisioning has become one of the central issues of 3G/4G mobile network design and analysis. Mobile multimedia applications have high demands in terms of available network resources, equipment design and QoS performances. The challenging task for research activities is the creation of adaptive application QoS protocol suite in order to obtain optimal QoS performance. In this paper we identify key adaptation issues and propose the architectural framework for application level QoS adaptation.

KEY WORDS: Mobile Networks, Quality of Service, Application, Protocols, Performance evaluation, Optimization

1. INTRODUCTION

The importance of Quality of Service (QoS) provisioning has become one of the central issues of 3G/4G mobile network design and analysis. The implementation of QoS in the actual network must be established on the "end-to-end" basis and must provide the service performance levels necessary for obtaining required Quality of Experience (QoE) for the end-user. Mobile multimedia applications have high demands in terms of available network resources, equipment design and QoS performances. The efficient employment of real-time applications in resource-constrained cellular environment remains an open field for research and analysis. Thus, the creation of an adaptive protocol suite with the possibility to adapt to dynamic network changes and diverse user requirements represents a challenging task.

2. PROBLEM FORMULATION

The variety of new applications (high definition video telephony, TV program distribution, tele-engineering, medical applications, location-based services, etc.) deployed in modern mobile networks implies the necessity for precise and consistent analysis. Each of these particular applications has its specific protocol stack with different parameter and attribute settings. Based on these settings, the QoS profile needs to be established and transmitted. Diverse applications share the same network resources and accordingly each service should be represented by properly chosen performance metrics. The common set of QoS parameters involved in the process of adaptation includes bandwidth, throughput, packet delay variation (jitter) and packet loss/error rate.

Most of the research activities have been oriented to the areas of system, network, transport and middleware support for QoS. The QoS mechanisms of the network and transport layer have been studied in depth, as well as resource reservation, admission control, routing, handoff procedures, service disciplines and traffic models.

Traditional transport protocols like TCP and UDP cannot support in entirety the QoS requirements of new mobile multimedia applications. Besides the effort to adapt and extent the usage of these protocols in cellular environment, other protocols have also been considered, such as DCCP (Datagram Congestion Control Protocol), SCTP (Stream Control Transport Protocol), etc.

The proposed architectures with adaptive mechanisms are provided mainly at the level of QoS signaling [1]. On the other side, there is a lack of proposed architectures and protocols for implementing an overall adaptive application QoS support. There are several ongoing research activities, but a common standard has not been adopted yet.

3. RELATED RESEARCH WORK

The research in this area is trying to address the key requirements needed for support of coordinated activities between application and critical system elements and resources. The general framework for QoS adaptation can be divided into two broad groups: network level and application level QoS adaptation. In both cases, the adaptation process follows the principle of formal control theory [2]. This approach has been used as a framework for the concept of "network-aware" applications [3]. The application has to adapt to changes in the available QoS while the network has to adapt to the alternations made to the QoS requirements of the supported applications.

As a direction for further research, some general recommendations related to QoS parameterization are emphasized. One approach for QoS provisioning is based on a combination of converters and description syntax that specifies QoS at the end-user, network and application level [5]. An adaptive QoS framework for integrated cellular and WLAN networks was proposed by [6]. The model supports the delivery of adaptive real-time flows using the QoS reservation-based approach.

Figure.1. General end-to-end reference model for QoS interworking (TR 23.802) [4]

The document 3GPP TS 23.802 "Architectural Enhancements for End-to-End QoS" considers possible solutions to enhance the "end-to-end" QoS architecture and to enable improved "end-to-end" QoS in the case of interworking with IP network domains [4]. The general "end-to-end" reference model for QoS interworking is shown in Figure 1. This model gives a brief overview of some application-related tasks in a heterogeneous network infrastructure.

Application nodes represent the interface between the domain specific nodes and backbone network. SIP (Session Initiation Protocol) has been chosen as the application control protocol.

The EU QoS project has the objective to explore QoS technologies for the advanced QoS-aware applications (voice, video-conferencing, video-streaming, educational, tele-engineering and medical applications) over various network infrastructures and domains [7].

In order to support QoS requirements for multimedia group communications in B3G networks, the QoS Architecture for Mobile Multicast Multimedia Services (Q3M) has been developed [8]. This architecture is oriented towards the establishment of adaptive multicast in DiffServ (Differentiated Services) mobile environments, with the seamless mobility support for streaming multimedia applications.

The rate adjusting and adaptation process at the application level incorporate several mechanisms: controlled quality degradation, adaptation of data format, implementation of seamless handoff procedures, as well as high QoS performances with low jitter, delay, low packet loss and guaranteed bandwidth. To achieve these requirements, different studies have been performed addressing the following methods: layered and multiple encoding, payload compression, correction, buffering, etc. [9].

4. APPLICATION LEVEL QOS

One of the key requirements for establishing the architecture of particular service is the creation of adequate framework for QoS performance analysis. The selection of appropriate application QoS protocol as well as mechanisms for its adaptation and optimization represents a broad field for research in today's mobile networks. The optimization process at the application level should be applied on media components, because of aggregation of heterogeneous traffic types.

It is very important to select an adequate performance metrics for particular application [10]. Applications can differ to a great extent when comparing the metrics for performance evaluation. For example, the performance metrics for real time video sharing application are the set-up delay for establishing connection and the delay experienced during the period of recording video on the transmitting side until receiving the content on the other end. For a similar application, video streaming, the low delay is not of vital importance, but the high bit rate available for transmission. The adaptation approach for streaming multimedia services is oriented towards using buffering and transcoding proxies.

The implementation of adaptation module at the application level has important advantages in comparison to implementation on lower layers. The lower levels are already involved in numerous system tasks and perform many functions. As a general rule, the adaptation procedures on lower layers are implemented in hardware elements.

The adaptation at the application layer can be implemented as a software module which makes it easier for changes and restructuring. Moreover, the question is how many functions and processes can be performed at the application level, so the lower levels may perform other tasks. In that way the overall processing performances can be improved.

5. PROPOSED FRAMEWORK FOR QOS ADAPTATION

The idea behind the proposed adaptation framework is the creation of QoS profile transparent to lower layers, particularly the network layer. This profile is established as a set of entities, where each of them performs specific tasks, as shown in Figure 2.

5.1. Traffic characterization entity

The aggregation of different traffic types makes the problem of creating platform for QoS adaptation even more complex. Traffic modeling has great impact on obtaining optimal QoS performances. These profile information will be mapped onto the set of traffic representatives. The example of QoS test/reference profile is shown in Table 1. Applications with similar requirements can be aggregated to form one profile, where different traffic classes are based on the representative QoS attributes set.

Table 1. An example of test profile

For example, videoconferencing is one of the most challenging applications to support since it requires high bandwidth, low latency and low jitter. Depending on chosen video format (frame rate, frame size, color depth, etc.), codec type (H.263, MPEG-4, H.264, etc.) and traffic distribution parameters, the QoS requirements may differ significantly. The video traffic shows long-range dependency effects, and it is of significant importance to employ high quality traffic model, especially for simulation purposes.

5.1.1. Simulation results for videoconference application

For obtaining some QoS performances based on simulation of the videoconferencing application, we used the industry-standard network simulation tool, OPNET Modeler [11]. Our goal was to illustrate the complexity in modelling real-time application such as videoconferencing, and to show the necessity for including adaptation mechanisms. The simulation setup is shown in Figure 3.

Figure 2. Application-layer QoS adaptation framework

The videoconference traffic is transmitted between two nodes inside the UMTS network. We measured QoS performances by varying specific system and application parameters. The system parameters we included in our observation were background system utilization, throughput-based admission control, etc. For application parameters, we measured "end-to-end" videoconference delay and video jitter, by changing payload size, QoS class, distribution parameters, max available bit rate on uplink and downlink, etc. We were testing the OPNET built-in video conferencing model, but we also imported H.263 video trace from [12]. The ToS (Type of Service) chosen for simulation purpose was interactive multimedia and background.

Figure 3. Simulation setup for videoconference performance testing

The results obtained from Figure 4. and 5. show clearly that the possibility for adaptation of the traffic characterization profile, as well as the correct system behavior prediction are of vital importance in order to obtain optimal QoS performances. Different traffic profiles have different QoS performance results. The proper selection of application parameters is in close relation to the efficient use of network and system parameters.

5.2. Estimation entity

The generated test profile represents the input for the estimation entity. The main task of this entity is to estimate attributes critical for particular bearer service. The general bearer performance attributes are "end-to-end" packet transfer delay, "end-to-end" delay variation, throughput, and packet loss/error rate.

Based on test profile generated in previous entity, estimation entity provides minimum QoS requirements extracted from raw data describing transmission environment. These requirements represent some form of preliminary profile recommendation. For example, the recommendation may suggest the use of DCH (Dedicated Channel) for real-time video transmission channel, as well as it may suggest the codec replacement, header compression or the introduction of new QoS parameters.

Additionally, the recommendation may include the suggestion for suitable application-level control protocol and transport protocol. The estimation entity output is the preliminary profile that will be used for the adoption of appropriate application policy.

Figure 4. "End-to-end" delay for three different scenarios

Figure 5. Video conferencing packet delay variation for three different scenarios

5.3. Application policy entity

This entity is in charge of establishing adaptation rules that will be used for the correction of initial test profile attributes. Mechanisms involved in this entity provide policies to dynamically regulate the behavior of system components involved in obtaining the adaptation profile. Based on this, adaptation rules are selected.

The policy entity should establish appropriate performance metrics for the application QoS profile evaluation. The complexity factor can be added to the preliminary profile to indicate the "QoS-aware" status of the particular profile or specific QoS mode that can be attached to the particular profile. The profile with this status will have higher priority in further processing.

5.4. QoS syntax entity

QoS syntax entity implements the profile description in specification language that enables dynamic and transparent correspondence with user and network requirements. The policy represents the specification of method or action that meets service requirements. This entity should enable the decomposition of protocol functionality into components in a way that they can be reassembled dynamically.

6. CONCLUSION

With the introduction of the new packet-optimized radio technologies (High Speed Downlink/Uplink Packet Access - HSDPA, HSUPA), as well as new radio access network architecture (3GPP Long Term Evolution) in mobile networks, the variety of multimedia applications become available. The "end-to-end" QoS support has been recognized as one of the key requirements for successful employment of modern mobile networks.

Based on the above-mentioned facts and the reviewed literature and published scientific papers, it can be concluded that there is a need for an in-depth study of application level QoS protocols and mechanisms.The challenging research task is the creation of an adaptive application QoS protocol suit for obtaining optimal QoS performance. Particularly, the main issues are:

- how to create an adaptive application-level QoS set of parameters and attributes;

- how to achieve the optimal mapping between application QoS parameters and system components.

In this paper, we identify the key adaptation issues and propose the architectural framework for the application level QoS adaptation. The next step will be the implementation and simulation verification of the proposed framework.

Acknowledgments

This work has been supported by the Laboratory of Mobile Radio Communications, School of Electrical and Computer Engineering, National Technical University of Athens, Greece.

References

[1] T. Guenkova-Luy, A. Kassler, and D. Mandato: "End-to-End Quality of Service Coordination for Mobile Multimedia Applications", IEEE Journal of Selected Areas in Communications (JSAC), Vol. 22, No. 5, June 2004.

[2] J. Bolliger, and T. Gross: "A Framework-Based Approach to the Development of Network- Aware Applications", IEEE Transactions on Software Engineering, Vol. 24, 1998.

[3] J. Cao, K.M. McNeill, D. Zhang, and J.F. Nunamaker, Jr.: "An overview of network-aware applications for mobile multimedia delivery", Proceedings of the 37th Annual Hawaii International Conference on System Sciences, Track 9 , Jan. 2004.

[4] 3rd Generation Partnership Project: Technical Report 23.802, V1.2.0, "Architectural Enhancements for End-to-End QoS", Sept. 2005, http://www.3gpp.org/ftp/Specs/html-info/23802.htm

[5] A. Thomas, "Supplying legacy applications with QoS: a description syntax at application, end-user and network level", Proceedings of the Software Engineering and Applications, Track 374-059, 2002.

[6] X. G. Wang, G. Min, J. E. Mellor, K. Al-Begain, and L. Gua, "An adaptive QoS framework for integrated cellular and WLAN networks", Computer Networks, Vol.47, No.2, Feb. 2005, pp. 167-183.

[7] http://www.euqos.eu/

[8] Project: "Q3M - QoS Architecture for Mobile Multicast Multimedia Services", http://www.workingonweb.com/q3m/

[9] F. Houéto, and S. Pierre: "Quality of service and performance issues in multiservice networks subject to voice and video traffics", Journal of Computer Communications, Vol.28, No.4, March 2005, pp. 393–404.

[10] D. Soldani, and M. Li, R. Cuny: "QoS and QoE Management in UMTS Cellular Systems", John Wiley & Sons, Ltd, 2006.

[11] http://www.opnet.com, OPNET Technologies homepage.

[12] www.tkn.tu-berlin.de/research/trace/ltvt.html

Authors

Gordana Gardašević holds a PhD degree in electrical engineering obtained from the Faculty of Electrical Engineering in Banja Luka, Bosnia and Herzegovina, in 2008. She is currently the Assistant Professor at the Department of Telecommunications of this faculty. Professor Gardašević has published 36 papers in scientific journals, national and international symposia. Her research interests include distributed multimedia systems and applications, quality of service in 3G/4G mobile networks, cross-layer design and next generation networks architectures and applications.

Milojko Jevtovićgraduated and obtained the Master of Science degree in electrical engineering from the Faculty of Electrical Engineering in Belgrade, Serbia, and the PhD degree in electrical engineering from the Faculty of Electrical Engineering in Zagreb, Croatia. Professor Jevtović has published 9 books and more than 160 papers in scientific journals, national and international symposia. He received several awards for his scientific work and contribution. Professor Jevtović is the member of Yugoslav Engineering Academy (JINA), Belgrade, Serbia.

Philip Constantinou graduated in physics at the National University of Athens, Greece, in 1972, he became the Master of Applied Science in electrical engineering at the University of Ottawa, Ontario, Canada, in 1976, and he obtained the PhD degree in electrical engineering from Carleton University, Ottawa, Ontario, Canada, in 1983. From 1976 to 1979 he was with Telesat Canada. In 1980, he joined the Ministry of Communications in Ottawa, Canada. From 1984 to 1989 he was with the National Research Centre Demokritos in Athens, Greece, where he was involved on several research projects in the area of Mobile Communications. In 1989, he joined National Technical University of Athens where he is currently Professor. Prof. Constantinou is also the Director of Mobile Radio Communications Laboratory. His current research interests include personal communications, mobile satellite communications, and interference problems on digital communications systems.