学霸学习网 这下你爽了
当前位置:首页 >> >>

Multimedia QoS Adaptation for Inter-tech Roaming

Multimedia QoS Adaptation for Inter-tech Roaming Using Spatially Situated Multicast Groups
Cristian Hesselman, Henk Eertink and Arjan Peddemors Telematica Instituut, the Netherlands
{hesselman, eertink, peddemors}

1. Introduction
We have recently proposed an application-level QoS adaptation service for roaming between wireless networks that are based on different technologies [1] (‘inter-tech’ roaming [2]). The service is a central part of the content distribution platform that we have introduced in [3]. In this extended abstract, we briefly discuss the platform (Section 2), the adaptation service (Section 3) and its supporting handoff protocol (Section 4). We then outline the requirements that we put on the network infrastructure (Section 5) and describe the purpose and structure of our demo (Section 6). We conclude with a short summary of our work (Section 7).

As an example, consider the distribution of a TV channel to mobile clients. Figure 1 shows how our system decomposes the TV broadcast.
station.platinum (s.p) receivers city.platinum (c.p) receivers (c.g) receivers



P5 P4
D1 D1

D2 D2




Ms.p Xs

Mc.p Xc


client device C1 multicast group Ms.p Ei Encoder of type i


TV session

2. The Platform
Our platform supports the distribution of multimedia streams (e.g., a streamed TV channel) to a diverse set of mobile clients. The philosophy behind the platform is to find a balance between high scalability and the delivery of an optimal QoS to individual mobile clients. To accomplish this, the platform divides the coverage area of a wireless infrastructure into domains and restricts the amount of available ‘QoS spectrum’ in each domain to a few application-level service classes [4, 5]. A service class defines a domain-specific QoS level that the presentation resources (e.g., a display) of a mobile client receive. Each domain of a wireless infrastructure must support a few service classes and realize each class as a set of site-local multicast groups [6]. The size of this set may vary depending on whether the audio and video parts of a stream are encoded separately or jointly. It also depends on whether the encoding technique is layered [7]. We use proxies [4, 5, 8, 9] to bridge the differences between service classes. Our proxies connect to sets of site-local multicast groups for communications with mobile clients, and to a global multicast group to communicate with fixed clients. Proxies perform functions such as rate adaptation, transcoding, audio and video filtering, and so on. Proxies typically run on gateway hosts in an access domain [5, 9].


Decoder of type i

Figure 1. Example: a TV broadcast. The server-side application S produces an uncompressed audio-video stream and injects it into a TV session (dashed line). The session delivers the stream to client-side application components P1 through P7. Each Pi represents the presentation resources of client device Ci and forms the sink of the TV stream. Clients C1 through C7 mostly consist of mobile devices that are distributed over domains and The city domain operates a medium-range medium-speed Metropolitan Area Network (MAN). It overlays [10, 11] the short-range high-speed LAN that operates. Each client device is equipped with at least two network interfaces so that they can connect to’s LAN as well as to’s MAN. The clients are furthermore IP multicast enabled and are equipped with an RTP depacketizer, a decoder (e.g., an MPEG-4 decoder) and multimedia presentation resources (display and speakers). All clients thus have the necessary resources for their end-users to watch the TV channel transmitted by S, regardless of whether they are in or To simplify the discussion, we assume that all the client devices demand that the audio and video portions of the

TV stream are encoded together in a non-layered fashion. As a result, the station and city domains can realize their service classes as a single site-local multicast group. For ease of notation, we denote a service class as domainName.className. We denote the multicast group associated with domainName.className as Md.c with d the first letter of domainName and c the first letter of className. We depict each Md.c as a multi-point arrow. As an example, consider station.platinum and its associated multicast group Ms.p. Players P1 through P3 receive the TV stream at class station.platinum because their respective client devices (C1 through C3) have joined Ms.p on their LAN interfaces. The proxy of (Xs) has joined the global multicast group as well as Ms.p. It realizes the QoS level associated with station.platinum by manipulating the media stream coming from S and transmitting it onto Ms.p. Xs may for instance have to adapt the rate of the TV stream to that of class station.platinum.

3. Inter-tech Roaming
The clients involved in the TV session of Figure 1 will typically experience fluctuations in the availability of communications resources [12]. As a result, the QoS of the stream that the players receive needs to be adapted [13] so that the communications resources that are available can support it. Our platform adapts the QoS of the TV stream that a player receives by transferring the player to another service class. We call this a service class handoff. As an example, consider player P3 of Figure 1 and assume that the client device that hosts it (C3) roams from to Figure 2 shows the portion of Figure 1 that is relevant for this particular situation.
station.platinum (s.p) receivers city.platinum (c.p) receivers

platform furthermore configures C3’s decoder and IP service so that their QoS characteristics correspond to the QoS level that city.platinum defines. Observe that in this particular example city.platinum is a suitable target class because it is also based on encoding type 1. The result of the handoff is that P3 receives the TV stream from Xc rather than from Xs. In addition, it receives the stream at the QoS level associated with class city.platinum. This QoS level will typically be lower than that of station.platinum because uses a relatively low-bandwidth MAN rather than the highbandwidth LAN of The end-user will therefore perceive a degradation in the QoS of the audiovideo stream that P3 presents when it moves from to The QoS will usually improve when C3 roams back into the station domain. The networks that the station and city domains use will most likely be based on different technologies. Using the terminology of Pahlavan et al. [2], we therefore call the handoff of Figure 2 an inter-tech service class handoff. In this example, it is also an inter-domain [9] service class handoff because it transfers P3 from the station to the city domain. Service class handoffs can also occur when a client roams in a network based on a single technology (intra-tech) or within a single domain (intra-domain). Combinations of these types of handoffs are also possible.

4. Inter-tech Handoff Protocol
The inter-tech service class handoffs that our platform supports are mobile-controlled [11]. In the example of Figure 2, this means that there is a handoff component on C3 that decides whether a handoff to Mc.p is required. The handoff component is part of our platform. It also executes a handoff should one be required. Our handoff component uses the packet loss characteristics of the paths between C3 and the two proxies to decide when to handoff to another service class. We make use of beaconing [10] to determine these loss characteristics. In the example of Figure 2, this means that proxies Xs and Xc multicast beacon messages into their domains at regular intervals. Xs and Xc include the domain that they belong to and the beacon interval that they use in the beacon messages. Both proxies transmit their beacon messages onto the same well-known multicast group Mb. Mb is a site-local multicast group in both and The beacons that Xs and Xc transmit do therefore not leave their respective domains. The handoff component uses the beacon messages from Xs and Xc to determine if C3 should use its LAN interface or its MAN interface to receive the TV stream. The handoff component joins C3 to Ms.p (class station.platinum) if it decides to use the LAN interface; it joins C3 to Mc.p (class city.platinum) if it decides to use C3’s MAN interface.


D1 D1





Ms.p Xs

Mc.p Xc TV session


Encoder of type i


Decoder of type i

Figure 2. A service class handoff. Since service classes are domain-specific, our platform must eventually handoff player P3 from class station.platinum to class city.platinum. To achieve this, the platform unsubscribes C3 from the multicast group associated with station.platinum (Ms.p) and joins it to the multicast group of city.platinum (Mc.p). The

6.1 Proxies

5. Infrastructure Requirements
Our approach requires a server in the network infrastructure that hands out IP addresses, for instance a DHCP [14] server. A mobile client must receive an IP address from this server whenever it roams into a new network. For example, C3 needs to receive an IP address from the server when it roams from the MAN of into the LAN of C3 can join Mb and Ms.p only after it has associated the new address with its LAN interface. In our approach, the IP address of a mobile client changes regularly as a result of roaming. This is similar to the SIP solution for mobility [15]. Our model does not require a client to have a permanent IP address like in Mobile IP [16]. As a result, we also do not require any additional routing complexity from the IP layer, i.e. we do not need Mobile IP. We further require base stations to be connected to a multicast router through a separate link or through a multicast-aware switch. To see why this is necessary, consider clients C4 through C7 of Figure 1. Suppose that C4 and C5 reside in one cell (A) and C6 and C7 in another (B). If the base stations of both cells are connected to the local multicast router through a shared link that does not involve a multicast-aware switch, the traffic of Mc.g would be transmitted into cell A while there are no gold clients in that cell. Similarly, the traffic of Mc.p would be sent into cell B while there are no platinum clients in that cell. This is clearly an undesirable situation. The Solaris machine in Figure 3 hosts the proxies Xs and Xc of Figure 2. For reasons of simplicity, we have implemented proxies Xs and Xc to act as broadcasting servers. That is, they generate the stream containing the TV channel locally rather than from a stream coming from the broadcasting server S (cf. Figure 1). Xs and Xc each consist of a QuickTime Darwin streaming server [17]: Xs consists of server Ss; Xc consists of server Sc. Ss and Sc run synchronously as indicated by the arrow between them in Figure 3 and loop continuously. Ss locally reads a high quality movie from a hinted (i.e., encoded and RTP-packetized) QuickTime file and transmits it onto the multicast group that represents class station.platinum, Ms.p. Similarly, Sc locally reads a low quality version of the same movie from a different hinted file and transmits it onto the multicast group that represents class city.platinum, Mc.p. Xs and Xc each also contain a process (not shown in Figure 3) that transmits beacon messages at regular intervals.

6.2 Networks
The RadioLan [18] base station in Figure 3 represents’s LAN (high capacity, short range). It offers a

6. Demo
Our demo implements the scenario of Figure 2 in which client C3 roams between domains and Figure 3 shows its organization. This figure also illustrates how the proxy and player components of Figure 2 are distributed over the various machines that the demo is made up of.
Win98 Laptop (C3) “” “” Ms.p RadioLan base station

gross bandwidth of 10 Mbps. The RadioLan base station provides an indoor range of approximately 15 meters and operates in the 5.8 GHz band. The WaveLan [19] base station mimics the MAN of (medium capacity, medium range). It provides a gross over-the-air bandwidth of 1 Mbps. The WaveLan base station has an indoor range of approximately 30 meters and operates at a frequency of 2.4 GHz. We have placed the RadioLan and WaveLan base stations next to each other. In this way, the WaveLan cell overlays the RadioLan cell.

6.3 Client
We use a Windows98 laptop to represent client C3. The laptop is equipped with a RadioLan and a WaveLan network interface that have hard-coded IP addresses. The RadioLan interface represents C3’s LAN interface; the WaveLan interface represents its MAN interface. The laptop runs the QuickTime client software package [20] to receive the streams that servers Ss and Sc transmit. It hosts a Java implementation of the handoff component. The handoff component uses the beacons that it receives from Xs and Xc to determine the network that the laptop should use (RadioLan or WaveLan). It also executes a handoff by calling the QuickTime API. For example, if the laptop roams out the RadioLan coverage area (see Figure 3), the handoff component observes that it begins to loose beacons from Xs. At some point, it will decide that it has

WaveLan base station

Xs Ss


Sc Xc

Linux PC

Solaris server

Figure 3. Demo.

lost too many beacons and call the QuickTime API to join the laptop to Mc.p on its WaveLan interface. When the handoff is completed, the end-user sees the low quality version of the stream that Sc produces rather than the stream from Ss (see Figure 3). Figure 4 shows this posthandoff situation.
Win98 Laptop (C3) “” “” Ms.p RadioLan base station WaveLan base station



Xs Ss


Sc Xc

Linux PC

Solaris server

Figure 4. Post-handoff situation.

7. Summary
We introduced a platform that revolves around the notion of domain-specific application-level service classes. We focussed on the QoS adaptation service of our platform, in particular for inter-tech roaming scenarios. The service uses beacon messages that service providers transmit to realize inter-tech roaming. Mobile clients listen for these beacons and use them to determine when they should initiate and complete a handoff to a new service class. We used IP multicast to increase the scalability of our platform. We feel that one of the major advantages of our approach is that it handles mobility at the application level. This allows our platform to deal with handoffs in an application-aware manner. In addition, we believe that our approach makes the selection and adaptation of QoS levels relatively simple. Similar to the SIP solution for mobility, we think that our approach would typically be used in conjunction with Mobile IP. Our approach would support applications that use UDP while Mobile IP would support TCP-based applications. A disadvantage of our approach is that a mobile client does not have a unique identifier. As a result, we will have to deal with authentication at the application-level rather than at the IP level (cf. Mobile IP).

[2] K. Pahlavan, P. Krishnamurthy, A. Hatami, M. Ylianttila, J. Makela, R. Pichna, J. Vallstr?m, “Handoff in Hybrid Mobile Data Networks”, IEEE Personal Communications, April 2000 [3] C. Hesselman, H. Eertink, “A Scalable QoS Adaptation Service for Mobile Multimedia Applications”, 6th EUNICE Open European Summer School, Sept. 2000, Enschede, The Netherlands, 4-4.pdf [4] G. Fankhauser, M. Dasen, N. Weiler, B. Plattner, B. Stiller, “The WaveVideo System and Network Architecture: Design and Implementation”, Technical Report, Zürich, Swizterland, June 1998 [5] N. Yeadon, F. Garcia, D. Hutshison, D. Shepherd, “Filters: QoS Support Mechanisms for Multipeer Communications”, IEEE Journal on Selected Areas in Communications, Sept. 1996 [6] D. Lee, D. Lough, S. Midkiff, N. Davis, P. Benchoff, “The Next Generation of the Internet: Aspects of the Internet Protocol Version 6”, IEEE Network, Jan/Feb 1998 [7] H. Schulzrinne, S. Casner, R. Frederick, V. Jacobson, “RTP: A Transport Protocol for Real-Time Applications”, Internet draft, draft-ietf-avt-rtp-new-08.txt, July 2000 [8] I. Kouvelas, V. Hardman, J. Crowcroft, “Network Adaptive Continuous-Media Applications Through Self Organised Transcoding”, Proc. of Network and Operating Systems Support for Digital Audio and Video (NOSSDAV’98), July 1998, Cambridge, UK [9] A. Balachandran, A. Campbell, M. Kounavis, “Active Filters: Delivering Scaled Media to Mobile Devices”, 7th Int. Workshop on Network and Operating System Support for Digital Audio and Video (NOSSDAV’97), St. Louis, USA, May 1997 [10] M. Stemm, R. Katz, “Vertical Handoffs in Wireless Overlay Networks”, ACM Mobile Networking, Special Issue on Mobile Networking and Internet, Spring 1998 [11] N. Tripathi, J. Reed, H. VanLandingham, “Handoff in Cellular Systems”, IEEE Personal Communications, Dec. 1998 [12] A. Campbell, “A Research Agenda for QOS-aware Mobile Multimedia Middleware”, NSF Wireless and Mobile Communications Workshop, Virginia, USA, March 1997 [13] D. Chalmers, M. Sloman, “A Survey of Quality of Service in Mobile Computing Environments”, IEEE Communications Surveys,, 2nd Quarter 1999 [14] R. Droms, “Automated Configuration of TCP/IP with DHCP”, IEEE Internet Computing, July-August 1999 [15] E. Wedlund, H. Schulzrinne, “Mobility Support Using SIP”, 2nd ACM/IEEE International Conference on Wireless and Mobile Multimedia (WoWMoM’99), Seattle, USA, Aug. 1999 [16] J. Solomon, “Mobile IP — The Internet Unplugged”, Prentice Hall, 1998 [17] Darwin Streaming Server, [18] RadioLan homepage, [19] WaveLan homepage, [20] Apple QuickTime Client 4.1,

8. References
[1] C. Hesselman, H. Eertink, A. Peddemors, “Multimedia QoS Adaptation for Inter-tech Roaming”, paper submitted to the IEEE Symposium on Computers and Communications (ISCC’01), Tunisia, July 2001


Multimedia QoS Adaptation for Inter-tech Roaming.pdf

Multimedia QoS Adaptation for Inter-tech Roaming_专业资料。We have recently proposed an application-level QoS Multimedia QoS Adaptation for Inter-tech Roaming...

Multimedia QoS adaptation for inter-tech roaming.unkown

// Multimedia QoS adaptation for inter-tech roaming CONFERENCE PAPER in PROCEEDINGS - INTERNATIONAL SYMPOSIUM ON ...

Multimedia QoS Adaptation for Inter-tech Roaming.unkown

Multimedia QoS Adaptation for Inter-tech Roaming Cristian Hesselman, Henk Eertink and Arjan Peddemors Telematica Instituut, The Netherlands {hesselman, eer...

Multimedia QoS Adaptation for Inter-tech Roaming Using ....unkown

an application-level QoS adaptation service for roaming between wireless networks that are based on different technologies [1] ('inter-tech' roaming [2])...


802.11e QoS Duncan Kitchin Wireless LAN Architect Mobile Communications ...roaming 802.11e quality of service 802.11f inter-access point ...

...of a Mobile QoS Adaptation Strategy for Wireless ATM.pdf

1 Network QoS Adaptation ‘‘Roaming M-QoS’’ Reference Application QoS ...The proposed M-QoS refresh of hando s, i.e., intra and inter. and ...


We can even create “yellow pages” directories for roaming ad hoc devices... resource adaptation should be supported by intergrid QoS adaptation services ...

...Bandwidth Allocation Algorithm in the Wireless ATM for a ....pdf

need guarantees on the QoS (Quality of Service...roaming (the localization of the mobile) and the... in WATM by an adaptation of the BAAWAM ...

Broadcasting Multimedia Channels in Future Mobile Systems_....pdf

Broadcasting Multimedia Channels in Future Mobile Systems_专业资料。Abstract. ...Peddemors, “Multimedia QoS Adaptation for Inter-tech Roaming”, Proc. of ...

Conceptual Architecture for QoS in Heterogeneous Networks_....pdf

which support global roaming across multiple wireless and mobile networks [3...end congestion control in networked multimedia applications with QoS requirements...

End to End QoS Platform CONTENT_图文.pdf

End to End QoS Platform CONTENT_专业资料。Fixed ...Easy adaptation to New Service and Technologies ...Roaming testing to analyze self performance and ...

Multimedia QoS Adaptation for Inter-tech Roaming Using ....unkown

// Multimedia QoS Adaptation for Inter-tech Roaming Using Spatially Situated Multicast Groups ARTICLE CITATION 1 ...

3G通信协议算法 物理层算法的研究、仿真和设计 链路仿真 代码实现....doc

adaptation of the radio transmission to the Internet packet traffic flows and...(QoS) to a wide range of applications while allowing seamless roaming among...

A Survey of Routing Protocols that Support QoS in Mobile Ad ....unkown

inter- ference, and shadowing; causing unpredictable link band- width and about network resource availability, can benefit from QoS adaptation....

Task-based Adaptation for Ubiquitous Computing.unkown

Focusing on the engineering issues of self-adaptation in the presence of heterogeneous plat- forms, legacy applications, mobile users, and resource variable ...

Dynamic QoS Adaptation for Multimedia Traffic.unkown

Dynamic QoS Adaptation for Multimedia Traffic Laercio Cruvinel, Teresa Va... but interoperating, models for the core and for the edge of the network...

...Cross Layer QoS Adaptation Architecture for Multimedia ....unkown

(0975 8887) Volume 64 No.8, February 2013 MPEG-21 Integrated Cross Layer QoS Adaptation Architecture for Multimedia Applications in Future MANETs to...

Technical Review.unkown

roaming access is so expensive that it would ...(DAB) Dynamic Label Segment Digital Multimedia ...(i.e. packets) for QoS, channel change, or ...

QoS Adaptation in Multimedia Multicast Conference ....unkown

(WWW) service has incorporated new distributed multimedia conference applications...QoS adaptation, when treated systematically in tolerant real-time applications,...

网站首页 | 网站地图
All rights reserved Powered by 学霸学习网
copyright ©right 2010-2021。