Document not found! Please try again

End-to-End QoS for Video Delivery Over Wireless Internet

11 downloads 65738 Views 425KB Size Report
proaches in QoS support: the network-centric and the end-system centric solutions. ... management, business models, and finally the main target, end users. ...... Prior to his current post, he was with Bell Labs., Lucent Technologies,. Murray Hill ...
End-to-End QoS for Video Delivery Over Wireless Internet QIAN ZHANG, SENIOR MEMBER, IEEE, WENWU ZHU, SENIOR MEMBER, IEEE, YA-QIN ZHANG, FELLOW, IEEE

AND

Invited Paper

Providing end-to-end quality of service (QoS) support is essential for video delivery over the next-generation wireless Internet. In this paper, we address several key elements in the end-to-end QoS support, including scalable video representation, network-aware end system, and network QoS provisioning. There are generally two approaches in QoS support: the network-centric and the end-system centric solutions. The fundamental problem in a network-centric solution is how to map QoS criterion at different layers respectively, and optimize total quality across these layers. In this paper, we first present the general framework of a cross-layer network-centric solution, and then describe the recent advances in network modeling, QoS mapping, and QoS adaptation. The key targets in end-system centric approach are network adaptation and media adaptation. In this paper, we present a general framework of the end-system centric solution and investigate the recent developments. Specifically, for network adaptation, we review the available bandwidth estimation and efficient video transport protocol; for media adaptation, we describe the advances in error control, power control, and corresponding bit allocation. Finally, we highlight several advanced research directions. Keywords—Cross-layer, end-system centric, end-to-end QoS, network-centric, video delivery, wireless Internet.

I. INTRODUCTION With the rapid growth of wireless networks and great success of Internet video, wireless video services are expected to be widely deployed in the near future. As different types of wireless networks are converging into all IP networks, i.e., the Internet, it is important to study video delivery over the wireless Internet. The current trends in the development of real-time Internet applications and the rapid growth of mobile systems indicate that the future Internet architecture will need to support various applications with different

Manuscript received January 16, 2004; revised July 20, 2004. The authors are with the Beijing Sigma Center, Microsoft Research Asia, Beijing 100080, China (e-mail: [email protected]; wwzhu@ microsoft.com; [email protected]). Digital Object Identifier 10.1109/JPROC.2004.839603

quality of service (QoS)1 requirements [1]. QoS support is a multidisciplinary topic involving several areas, ranging from applications, terminals, networking architectures to network management, business models, and finally the main target, end users. Enabling QoS in Internet is difficult, and becomes more challenging when introducing QoS in an environment involving mobile hosts under different wireless access technologies, since available resources (e.g., bandwidth, battery life, etc.) in wireless networks are scarce and dynamically change over time. For wireless networks, since the capacity of a wireless channel varies randomly with time, providing deterministic QoS (i.e., zero QoS violation probability) will likely result in extremely conservative guarantees and waste of resources, which is hardly useful. Thus, in this paper, we only consider statistical QoS [3]. To support end-to-end QoS for video delivery over wireless Internet, there are several fundamental challenges. 1) QoS support encompasses a wide range of technological aspects. To be specific, many technologies, including video coding, high-performance physical and link layers support, efficient packet delivery, congestion control, error control, and power control, all affect the overall QoS. 2) Different applications have very diverse QoS requirements in terms of data rates, delay bounds, and packet loss probabilities. For example, unlike nonreal-time data packets, video services are very sensitive to packet delivery delay but can tolerate some frame losses and transmission errors. 3) Different types of networks inherently have different characteristics. This is also referred to as network heterogeneity. It is well known that Internet is based on Internet Protocol (IP), which basically only offers the 1Note that the definition of QoS in itself may be somewhat confusing and has different implications. We adopt the definition “the ability to ensure the quality of the end user experience” [2] in this paper.

0018-9219/$20.00 © 2005 IEEE

PROCEEDINGS OF THE IEEE, VOL. 93, NO. 1, JANUARY 2005

123

Fig. 1.

Fundamental components for end-to-end QoS support.

best-effort services. Specifically, network conditions, such as bandwidth, packet loss ratio, delay, and delay jitter, vary from time to time. An important characteristic of wireless networks in the future is that there are mixtures of heterogeneous wireless access technologies co-existed such as wireless local area network (WLAN) access, 2.5G/3G cellular access, and Bluetooth. Bit-error rate (BER) in a wireless network is much higher than that in a wireline network. Moreover, link layer error control scheme, such as automatic repeat request (ARQ), is widely used to overcome the varying wireless channel errors. This will further increase the dramatic variation of bandwidth and delay in wireless networks. To make things even more complicated, the end-to-end packet loss in wireless Internet can be caused by either congestion loss occurred due to buffer overflow or the erroneous loss occurred in the wireless link due to channel error. 4) There is dramatic heterogeneity among end users. End users have different requirements in terms of latency, video visual quality, processing capabilities, power, and bandwidth. It is thus a challenge to design a delivery mechanism that not only achieves efficiency in network bandwidth but also meets the heterogeneous requirements of the end users. To address the above challenges, one should support the QoS requirement in all components of the video delivery system from end to end, which include QoS provisioning from networks, scalable video presentation from applications, and network adaptive congestion/error/power control in end systems. Fig. 1 illustrates key components for end-to-end QoS support. • QoS provisioning from networks. The best-effort nature of Internet has promoted the Internet Engineering Task Force (IETF) community to seek for QoS support through network layer mechanisms. The most well-known mechanisms are the Integrated Services (IntServ) [4] and the Differentiated Services (DiffServ) [5]. The approaches to providing QoS in wireless networks are quite different from their Internet counterparts. General Packet Radio Service (GPRS)/Universal Mobile Telecommunications System (UMTS) and IEEE 802.11 have total different mechanisms for QoS support. 124

• Multilayered scalable video coding from applications. In scalable coding, the signal is separated into multiple layers of different visual importance. The base layer can be independently decoded and it provides basic video quality. The enhancement layers can only be decoded together with the base layer and they further refine the video quality. Enhancements on layered scalable coding have proposed to provide further fine granularity scalability [7], [8], [95]. Scalable video representation provides fast adaptation to bandwidth variations as well as inherent error resilience and complexity scalability properties that are essential for efficient transmission over error prone wireless networks. • Network adaptive congestion/error/power control in end systems. When network condition changes, the end systems can employ adaptive control mechanisms to minimize the impact on user perceived quality. Power control, congestion control, and error control are three main mechanisms to support quality of services for robust video delivery over wireless Internet. Power control is performed collectively from the group point of view by controlling transmission power and spreading gain for a group of users so as to reduce interference [9]. Congestion control and error control are conducted from the individual user’s point of view to effectively combat the congestions and errors occurred during transmission by adjusting the transmission rate and allocating bits between source and channel coding [10], [11]. There have been two approaches in providing the end-to-end QoS support: the first one is network-centric QoS provisioning, in which routers/switches, or/and base stations/access points in the networks provide prioritized QoS support to satisfy data rate, delay bound, and packet loss requirements by different applications. In the prioritized transmission, QoS is expressed in terms of probability of buffer overflow and/or the probability of delay violation at the link layer. However, at the video application layer, QoS is measured by the mean squared error (MSE) and/or peak-signal-to-noise ratio (PSNR). Thus, one of the key issues for end-to-end QoS provisioning using network-centric solution is the effective QoS mapping across different layer. More specifically, one needs to consider how to model the varying network and coordinate effective adaptation of PROCEEDINGS OF THE IEEE, VOL. 93, NO. 1, JANUARY 2005

Fig. 2. General framework of end-to-end QoS support for video over wireless Internet with network-centric solution.

QoS parameters at video application layer and prioritized transmission system at link layer. In Section II, we will describe a general framework of a cross-layer architecture of a network-centric end-to-end QoS support solution and then review recent developments in individual components including network QoS support, channel modeling, QoS adaptation, and QoS mapping. The second type of approach to provide end-to-end QoS support is solely end-system centric. In particular, the end systems employ various control techniques, which include congestion control, error control, and power control, to maximize the application-layer video quality without any QoS support from the underlying network. The advantage of end system control is that there are minimum changes required in the core network. The main challenge, however, is how to design efficient power/congestion/error control mechanisms. In Section III, we will present a framework that targets at minimizing the end-to-end distortion or power consumption, and then review the recent studies on various mechanisms. II. NETWORK-CENTRIC CROSS-LAYER END-TO-END QoS SUPPORT As stated above, different layers (e.g., application layer and link/network layer) have different metrics to measure quality of service, which brings challenge for end-to-end QoS provisioning. Fig. 2 shows the general block diagram of end-to-end QoS support for video delivery in the network-centric cross-layer solution. This solution considers an end-to-end delivery system for a video source from the sender to the receiver, which includes source video encoding, cross-layer QoS mapping and adaptation, prioritized transmission control, adaptive network modeling, and video decoder/output modules. To support end-to-end QoS with network-centric approach, a dynamic QoS management

system is needed in order for video applications to interact with underlying prioritized transmission system to handle service degradation and resource constraint in time-varying wireless Internet. Specifically, to offer a good compromise between video quality and available transmission resource, the key is how to provide an effective cross-layer QoS mapping and an efficient adaptation mechanism. A. Network QoS Provisioning for Wireless Internet QoS provisioning for the Internet has been a very active area of research for many years. Two different approaches have been introduced in IETF, which are IntServ [4] and DiffServ [5], respectively. IntServ was introduced in IP networks in order to provide guaranteed and controlled services in addition to the existing best-effort service. IntServ and reservation protocols, such as ReSerVation Protocol (RSVP), have failed to become a practical end-to-end QoS solution for lack of scalability and difficulty in that all elements in the network have to be RSVP enable. DiffServ was proposed to provide a scalable and manageable network with service differentiation capability. In contrast to the per-flow-based QoS guarantee in the Intserv, Diffserv networks provide QoS assurance on a per-aggregate basis. The Internet research community has been proposing and investigating different approaches to achieve differentiated services. In particular, significant efforts have been devoted to achieve service differentiation in terms of queuing delay and packet loss [12], [13], both of which are of primary concern for multimedia applications. Many QoS control mechanisms, especially in the areas of packet scheduling [14], [15] and queue management algorithms [16], [17], have been proposed in recent years. Elegant theories, such as network calculus [18] and effective bandwidths [19], have also been developed. Firoiu et al. provided a comprehensive survey on a

ZHANG et al.: END-TO-END QoS FOR VIDEO DELIVERY OVER WIRELESS INTERNET

125

Fig. 3.

Different channel models.

number of recent advances in Internet QoS provisioning in [20]. There have also been many studies related to QoS provision in wireless networks. The Third Generation Partnership Project (3GPP)2 is the main standard body that defines and standardizes a common QoS framework for data services, particularly IP-based services. 3GPP has defined a comprehensive framework for end-to-end QoS covering all subsystems, from radio access network (RAN) through core network to gateway node (to the external packet data network) within a UMTS network [6]. 3GPP has also defined four different UMTS QoS classes according to delay sensitivity: conversational, streaming, interactive, and background classes. In wireless local area networks, the original IEEE 802.11 communication modes, namely, Distributed Coordination Function (DCF) and Point Coordination Function (PCF), do not differentiate traffic types. IEEE is proposing enhancements in 802.11e to both coordination modes to facilitate QoS support [21]. In Enhanced Distribution Coordination Function (EDCF), the concept of traffic categories is introduced. EDCF establishes a probabilistic priority mechanism to allocate bandwidth based on traffic categories. Aiming to extend the polling mechanism of PCF, Hybrid Coordination Function (HCF) is proposed. A hybrid controller polls stations during a contention-free period. The polling grants each station a specific start time and a maximum transmit duration. The 802.11e standard will be ratified at the end of this year. In the mean time, a group of vendors have proposed Wireless Multimedia Enhancements (WME) to provide an interim QoS solution for 802.11 networks [21]. WME uses four priority levels in negotiating communication between wireless access points and client devices. B. Cross-Layer QoS Support for Video Delivery Over Wireless Internet An efficient QoS mapping scheme that addresses cross-layer QoS issues for video delivery over wireless Internet includes the following important building blocks: 1) wireless network modeling that can effectively model 2www.3GPP.org

126

the time-varying and nonstationary behavior of the wireless networks; 2) prioritized transmission control scheme that can derive and adjust the rate constraint of a prioritized transmission system; and 3) QoS mapping and adaptation mechanism that can optimally map video application classes to statistical QoS guarantees of a prioritized transmission system so as to provide the best tradeoff between the video application quality and the transmission capability under time-varying wireless networks. 1) Wireless Network Modeling: One can model a communication channel at different layers, i.e., physical layer and link-layer (see Fig. 3). Physical layer channel can be further classified into radio-layer channel, modem-layer channel, and codec-layer channel. Among them, radio-layer channel models can be classified into large-scale path loss and small-scale fading [22]. Large-scale path loss models characterize the underlying physical mechanisms (i.e., reflection, diffraction, scattering) for specific paths. Small-scale fading models describe the characteristics of generic radio paths in a statistical fashion. Modem-layer channel can be modeled by a finite-state Markov chain [23], whose states are characterized by different BERs. A codec-layer channel can also be modeled by a finite-state Markov chain, whose states can be characterized by different data-rates, or a symbol being error-free/in-error, or a channel being good/bad [24]. Zorzi et al. [24] demonstrated that Markov model is an approximation on block transmission over a slowly fading wireless channel. In general, based on existing physical-layer channel models, it is very complex to characterize the relationship between the control parameters and the calculated QoS measures. This is because the physical-layer channel models do not explicitly characterize the wireless channel in terms of the link-level QoS metrics, such as data rate, delay, and delay violation probability. Recognizing that the limitation of physical-layer channel models in QoS support, i.e., the difficulty in analyzing linklevel performances, attempts have been made to move the channel model up in the protocol stack, from physical-layer to link-layer [25], [26]. In [25], an effective capacity (EC) channel model was proposed. The model captures the effect PROCEEDINGS OF THE IEEE, VOL. 93, NO. 1, JANUARY 2005

of channel fading for the link queueing behavior using a computationally simple yet accurate model, thus can be a critical tool for designing efficient QoS provisioning mechanisms. 2) Prioritized Transmission Control: To achieve differentiated services, a class-based buffering and scheduling mechanism is needed in the prioritized transmission control module. In particular, QoS priority classes are maintained with each class of traffic being maintained in separate buffers. Priority scheduling policy is employed to serve packets of the classes. Under this class-based buffering and priority scheduling mechanism, each QoS priority class can obtain a certain level of statistical QoS guarantees in terms of probability of packet loss and packet delay. Then, the next step is to translate the statistical QoS guarantees of multiple priority classes into rate constraints based on the effective capacity theory [25]. The calculated rate constraints in turn specify the maximum data rate that can be transmitted reliably with statistical QoS guarantee over the time-varying wireless channel. Consequently, video substreams can be classified into classes and bandwidth can be allocated accordingly for each class. The rate constraint of multiple priority classes under a time-varying service rate channel can be derived according to the guaranteed packet loss probabilities and different buffer sizes of each priority class [26]. The statistical QoS guarantee of each priority class is provided in terms of packet loss probability based on the effective service capacity theory. In [26], Kumwilaisak et al. derived the rate constraint of substreams under a simplest strict (nonpreemptive) priority scheduling policy. 3) QoS Mapping and QoS Adaptation: QoS mapping and QoS adaptation are the key components to achieve cross-layer QoS support in this video delivery architecture. Unlike the adaptive channel modeling module and prioritized transmission control module, the QoS mapping and QoS adaptation are application-specific. Since the QoS measure at the video application layer (e.g., distortion and uninterrupted video service perceived by end-users) is not directly related to QoS metrics at the link layer (e.g., packet loss/delay probability), a mapping and adaptation mechanism must be in place to more precisely match the QoS criterion across different layers. Specifically, at the video application layer, each video packet is characterized based on its loss and delay properties, which contributes to the end-to-end video quality and service. Then, these video packets are classified and optimally mapped to the classes of link transmission module under the rate constraint. The video application layer QoS and link-layer QoS are allowed to interact with each other and adapt to the wireless channel condition, whose objective is to find the QoS tradeoff, which simultaneously provides a desired video service of the end users with available transmission resources. There have been many studies on the cross-layer design for efficient multimedia delivery with QoS assurance over wired and wireless networks in recent years [13], [26]–[29]. The focus has been on the utilization of the differentiated service architecture to convey multimedia data. The common approach is to partition multimedia data into smaller units

and then map these units to different classes for prioritized transmission. The partitioned multimedia units are prioritized based on its contribution to the expected quality at the end user while the prioritized transmission system provides different QoS guarantees depending on its corresponding service priority. Servetto et al. [30] proposed an optimization framework to segment a variable bit rate source to several substreams that are transmitted in multiple priority classes. The objective is to minimize the expected distortion of the variable bit rate source. Shin et al. [13] proposed to prioritize each video packet based on its error propagation effect if it is lost. Video packets were mapped differently to transmission priority classes with the objective of maximizing the end-to-end video quality under the cost and/or price constraint. Tan et al. [28] examined the same problem as that formulated in [13] with different approaches for video prioritization. Other types of multimedia delivery over DiffServ network, such as prioritized speech and audio packets, were considered by Martin [31] and Sehgal et al. [27]. Considering the stochastic behavior of wireless networks, [32], [33] introduced a cross-layer design with adaptive QoS assurance for multimedia transmission where absolute QoS was considered. In [32], Xiao et al. studied the rate-delay tradeoff curve offered from the lower-layer protocol to the applications. Then, the application layer selected the operating point from this curve as a guaranteed QoS parameter for transmission. These curves are allowed to be changed as the wireless network environment changes. In [33], it investigated the dynamic QoS framework to adaptively adjust QoS parameters of the wireless network to match with time-varying wireless channel condition, in which the application was given the flexibility to adapt to the level of QoS provided by the network. Targeting at scalable video codec and considering the interaction between layers to obtain the operating QoS tradeoff points, in [26], the QoS mapping and adaptation for wireless network was addressed in the following two steps. First, find the optimal mapping policy from priority classes such that one GOP (group of picture) to the distortion of this GOP is minimized. Second, find a set of QoS parameters for the priority network, such that the expected video distortion is minimized. III. END-SYSTEM CENTRIC QoS SUPPORT To provide end-to-end QoS with end-system solution, the video applications should be aware of and adaptive to the variation of network condition in wireless Internet. This adaptation consists of network adaptation and media adaptation. The network adaptation refers to how many network resources (e.g., bandwidth and battery power) a video application should utilize for its video content, i.e., to design an adaptive media transport protocol for video delivery. The media adaptation controls the bit rate of the video stream based on the estimated available bandwidth and adjusts error and power control behaviors according to the varying wireless Internet conditions. The general diagram for end-system centric QoS provisioning is illustrated in Fig. 4. To address network adaptation, an end-to-end video transport protocol is needed to

ZHANG et al.: END-TO-END QoS FOR VIDEO DELIVERY OVER WIRELESS INTERNET

127

Fig. 4. General framework for end-to-end QoS provisioning for video over wireless Internet with end-system-centric solution.

handle congestion control in wireless Internet. More specifically, the Adaptive Network Monitor deals with probing and estimating the dynamic network conditions. The Congestion Control module adjusts sending rate based on the feedback information. For media adaptation, considering that different parts of compressed scalable video bitstream have different importance level, Network-aware Unequal Error Protection (UEP) module protects different layers of scalable video against congestive packet losses and erroneous losses according to their importance and network status. Network-aware Transmission Power Adjustment module adjusts the transmission power of the end-system to affect the wireless channel conditions. R-D Based Bit Allocation module performs media adaptation control with two different targets, i.e., distortionminimization and power consumption-minimization. A. Network Adaptive Congestion Control Bursty loss and excessive delay have a devastating effect on perceived video quality, and these are usually caused by network congestion. Thus, congestion-control mechanism at end systems is necessary to reduce packet loss and delay. Typically, for conferencing and streaming video, congestion control takes the form of rate control. Rate control attempts to minimize the possibility of network congestion by matching the rate of the video stream to the available network bandwidth. To deliver media content, several protocols are involved and some of them were proprietary solutions. Those protocols include the Real Time Transport Protocol (RTP) and Real Time Control Protocol (RTCP) [34], Session Description Protocol (SDP) [35], Real Time Streaming Protocol (RTSP) [36], Stream Control Transmission Protocol (SCTP) [37], Session Initiation Protocol (SIP) [38] and Hypertext Transport Protocol (HTTP). Since a dominant portion of today’s Internet traffic is TCP-based, it is very important for multimedia streams to be “TCP-friendly,” by which it means a media flow generates similar throughput as a typical TCP flow along the same 128

path under the same condition with lower latency. There are two existing types of TCP-friendly flow-control protocols for multimedia delivery applications: sender-based rate adjustment and model-based flow control. Sender-based rate adjustment [10], [39] performs additive increase and multiplicative decrease (AIMD) rate control in the sender as in TCP. The transmission rate is increased in a step-like fashion in the absence of packet loss and reduced multiplicatively when congestion is detected. This approach usually requires the receiver to send frequent feedback to detect congestion indications, which may potentially degrade the overall performance. Model-based flow control [40], [41], on the other hand, uses a stochastic TCP model [42], which represents the throughput of a TCP sender as a function of packet loss ratio and round trip time (RTT). One issue that should be considered for this type of approach is that the estimated packet loss ratio is not for the next time interval so as to affect the accuracy of the throughput calculation. While TCP-friendliness is a useful fairness criterion in today’s Internet, it is possible that future network architectures (in which TCP is either no longer the predominant transport protocol or has a very bad performance) will allow or require different definitions of fairness. For example, fairness definition for wireless networks is still subject to research since TCP performance in wireless networks is still need to be improved. Designing a transport protocol for video transmission over wireless Internet, several issues related to network condition estimation should be considered. The most important one is the estimation of congestion loss ratio. In wireless Internet, the end-to-end packet loss can be caused by either congestion loss due to buffer overflow or the erroneous loss occurred in the wireless link. Traditional TCP and TCP-friendly media transport protocols [43], [44] treat any lost packet as a signal of network congestion and adjust its transmission rate accordingly. However, this rate reduction is unnecessary if the packet loss is due to the error occurred in wireless link, which in turn causes bad performance for end-to-end delivery quality. The second issue is the round trip time (RTT) estimation. There is large variation in end-to-end delay in PROCEEDINGS OF THE IEEE, VOL. 93, NO. 1, JANUARY 2005

wireless Internet [52]. Sending only a single acknowledgment to measure the RTT during a predefined period of time may be inaccurate and fluctuate greatly. The third issue is the available bandwidth estimation. There are many studies on available bandwidth estimation in Internet, and how to apply those schemes for transport protocol design in wireless networks are now attracting much attention [44], [54]. 1) End-to-End Packet Loss Differentiation and Estimation: As stated above, the key issue of designing an efficient media transport protocol is to correctly detect whether the network is in congestion or not. Generally there are two types of methods to distinguish the network status [45], which are split connection and end-to-end method. In the split connection method, it requires an agent at the edge of wired and wireless network to measure the conditions of two types of networks separately [46], [47]. Specifically, an agent is needed at every base station in the entire wireless communication system, which adds excessive complexity in the actual deployment. The end-to-end method focuses on differentiating the congestive loss from the erroneous packet loss by adopting some heuristic methods such as interarrival time or packet pair [48]–[50]. This type of solution expects a packet to exhibit a certain behavior under wireless Internet. It is known that a specific behavior of a packet in the network reflects the joint effect of several factors. Considering that the traffic pattern in the Internet itself is a complicated research topic, finding a good pattern to predict the behaviors of packets in wireless Internet still requires some fundamental research. Yang et al. proposed a different mechanism in [51] to use the combined link layer and sequence number information to differentiate the wireless erroneous loss and congestive loss. The arrival time of the erroneous packets is used to derive the distribution of lost packets among the erroneous packets between two back-to-back correctly received packets. 2) Available Bandwidth Estimation: There are two types of approaches for available bandwidth estimation in media transport protocols. The first type of approach calculates the available bandwidth based on the estimated RTT and packet loss ratio. Padhye et al. [42] proposed a formula to calculate the network throughput that has been widely adopted [40], [52]. The second type of approach calculates the available bandwidth using the Receiver Based Packet Pair (RBPP) method [53]. RBPP requires the use of two consecutively sent packets to determine a bandwidth share sample. The most recognized scheme in this category is TCP-Westwood [54], which maintains two estimators, along with a method to identify the predominant cause of packet loss. Depending on the loss cause, the appropriate estimator is “adaptively” selected. One estimator, called Bandwidth Estimator (BE), considers each ACK pair separately to obtain a bandwidth sample, filters the samples, and returns to the (short term) bandwidth share that the TCP sender is getting from the network. The other estimator, called Rate Estimator (RE), measures the amount of data acknowledged during the latest interval . RE tends to estimate the (relatively longer term)

rate that the connection has recently experienced. Several media transport protocols, such as SMCC [55] and VTP [56], proposed recently, following the idea of TCP-Westwood. B. Adaptive Error Control There are two basic error correction mechanisms, namely, ARQ and FEC. ARQ has been shown to be more effective than FEC. However, FEC has been commonly suggested for real-time applications due to their strict delay requirements. Hybrid ARQ scheme proposed in [57] can achieve both delay bound and rate effectiveness by limiting the number of retransmissions. Other hybrid FEC and delay-constrained ARQ schemes were discussed in [58]–[60]. Girod and Färber reviewed on the existing solutions for combating wireless transmission errors in [61]. While their focus is on cellular networks, most presented protection strategies can also be applied to the transmission of video over other types of wireless networks. In [62], Shan and Zakhor presented an integrated application-layer packetization, scheduling, and protection strategies for wireless transmission of nonscalable coded video. Cote et al. presented a survey of the different video-optimized error resilience techniques that are necessary to accommodate the compressed video bitstreams [63]. Various channel/network errors can result in considerable damage to or loss of compressed video information during transmission. Effective error concealment strategies become vital for ensuring a high quality of the video sequences in the presence of errors/losses. A review of the existing error concealment mechanisms was given by Wang and Zhu in [64]. In [65], Majumdar et al. addressed the problem of resilient real-time video streaming over IEEE 802.11b WLANs for both unicast and multicast transmission. For the unicast scenario, a hybrid ARQ algorithm that efficiently combines FEC and ARQ was proposed. For the multicast case, progressive video coding based on MPEG-4 Fine Granularity Scalability (FGS) was combined with FEC. Scalable video has received lots of attention in recent years due to its fast adaptation characteristic. For scalable video, one way to efficiently combat channel errors is to employ unequal error protection (UEP) for information of different importance. More specifically, strong channel-coding protection is applied to the base layer data stream while weaker channel-coding protection is applied to the enhancement layer parts. Studying how to add FEC to scalable video coding has gained great interest recently. Joint work on scalable video coding with UEP for wired network [66], [67] and wireless communication [68]–[70] has been proposed. In [70], a network adaptive application-level error control scheme using hybrid UEP and delay constrained ARQ was proposed for scalable video delivery. Current and estimated round trip time is used at sender side to determine the maximum number of retransmission based on delay constraint. In [71], Van der Schaar and Radha discussed the combination of MPEG-4 FGS with scalable FEC for unicast and multicast applications, and a new unequal error protection strategy referred to as Fine Grained Loss Protection (FGLP) was introduced.

ZHANG et al.: END-TO-END QoS FOR VIDEO DELIVERY OVER WIRELESS INTERNET

129

Fig. 5. Illustration of rate-distortion with/without considering power constraint and transmission error.

It has been shown that under general wireless environments, different protection strategies exist at the various layers of the protocol stack, and hence a joint cross-layer consideration is desirable in order to provide an optimal overall performance for the transmission of video. A vertical system integration, referred to as “cross-layer protection,” was introduced in [72] that enabled the joint optimization of the various protection strategies existing in the protocol stack. Xu et al. developed a cross-layer protection strategy for maximizing the received video quality by dynamically selecting the optimal combination of application-layer FEC and MAC retransmission based on the channel conditions [73]. C. Joint Power Control and Error Control In general, there exists tradeoff between maintaining good quality of video application and reducing average power consumption, including processing power and transmission power at end-systems. From network point of view, multipath fading and multiple access interference (MAI) in wireless network necessitate the use of high transmission power. From video coding point of view, to decrease transmission power and maintain a desired video quality, more complex compression algorithms and more powerful channel coding schemes can be applied to source coding and channel coding, respectively. The motivation of jointly considering power control and error control for video communication comes from the following observations on the relationship among rate, distortion, and power consumption. Case According to the rate-distortion theory (Fig. 5, ), the lower the source coding rate , the larger the distortion . More generally, it can . be represented as When video compression is performed with Case a given power constraint , the power-constrained distortion includes both the distortion by the source rate control and the distortion ). caused by the power constraint (Fig. 5, More generally, it can be denoted as . Considering a more specific scenario, a video Case bitstream is transmitted over wireless links and a limited power with a given BER 130

constraint , the end-to-end distortion is composed of the distortion by the source rate control, the distortion caused by the channel errors, and the distortion caused by the power constraint (Fig. 5, ). More generally, it can . be denoted as From the individual user point of view, some studies on allocating available bits for source and channel coders are aiming at minimizing the total processing power consumption under a given bandwidth constraint. Specifically, a lowpower communication system for image transmission was investigated in [74]. A power-optimized joint source-channel coding (JSCC) approach for video communication over wireless channel was proposed in [75]. From the group user point of view, power control adjusts a group of users’ transmission powers to maintain their video quality requirements. Recently, the focus has been on adjusting transmission powers to maintain a required signal-tointerference ratio (SIR) for each network link using the least possible power. It is also referred to as resource management based on the power control technique discussed in [9], [76], [77], where it is formulated as a constrained optimization problem to minimize the total transmission power or maximize the total rate subject to the SIR and bandwidth requirements. The key observation Eisenberg et al. [78] and Zhang et al. [79] made independently is that when the transmission power of one user is changed to achieve its minimal power consumption, its interference to other users varies accordingly. This interference variation will alter other users’ receiving SIRs and may result in that their video quality requirements cannot be achieved, and then in turn deviate from the optimal state of their power consumptions. Therefore, due to the multiple access interference, the global minimization of power consumption must be investigated from the group point of view. D. Rate-Distortion Based Bit Allocation For video delivery over wired or wireless network, the most common metrics used to evaluate video quality are the expected end-to-end distortion and expected end-to-end . Here, consists of source distorpower consumption and channel distortion . The source distortion is tion caused by source coding such as quantization and rate control. The channel distortion occurs when the packet loss due to network congestion or wireless link error happened during consists of processing power on the the transmission. source coding , processing power on the channel coding , and the transmission power for data delivery . It is well known that channel bandwidth capacity is highly limited in wireless Internet. Thus, it is very important to efficiently allocate the bits among the source coding and the channel coding, under a given fixed bandwidth capacity so as to achieve the minimal expected end-to-end distortion or minimal expected end-to-end power consumption [67], [78]. More specifically, the resource allocation problem can be formulated as follows:

PROCEEDINGS OF THE IEEE, VOL. 93, NO. 1, JANUARY 2005

where is the total bandwidth assigned to source coding is the total bandwidth and channel protection, while budget. Or and where is the end-to-end distortion budget. In all the schemes mentioned above, the erroneous losses and the congestive losses are treated the same and only one type of packet loss is considered. As discussed earlier, in wireless Internet the packet losses consist of both congestive losses and erroneous losses, which in turn have different loss patterns in wireless and wired network parts. Considering that different loss patterns lead to different perceived QoS at application level [80], Yang et al. presented a loss differentiated – based bit allocation scheme [51], in which the channel distortion is caused by two parts: one is caused during the transmission over wired-line part of the connec, and the other is caused during the transmission tion, . over the wireless channel, IV. CONCLUSION In this paper, we reviewed recent advances in providing end-to-end QoS support for video delivery over wireless Internet from both network-centric and end-system centric perspectives. In the network-centric solution, we presented the general cross-layer QoS support architecture for video delivery over wireless Internet. This architecture enables one to perform QoS mapping between statistical QoS guarantees at the network level to a corresponding priority class with different video quality requirements. In the end-system centric approach, we described the framework that includes network adaptation and media adaptation and reviewed several key components in this framework. More specifically, recent developments in congestion control, error control, power control, and – based bit allocation schemes were addressed. Cross-layer design of heterogeneous wireless Internet video systems is a relatively new and active field of research, in which many issues need further examination. Optimally allocating resources in this heterogeneous setting presents many challenges and opportunities. To solve the cross-layer optimization problems for video transmission, several components such as: 1) adaptive modulation and channel coding; 2) adaptive retransmission; and 3) adaptive source rate control need to be jointly optimized to achieve better performance. Moreover, this paper is primarily focused on QoS support in a unicast scenario. Efficient end-to-end QoS support for multicast video transmission systems [81]–[84] is an area that still requires considerable work. Since it has been recognized that the Internet interdomain routing algorithm, Border Gateway Protocol (BGP), is not always able to provide good quality routes between domains, more recently, there have been proposals to establish application-level overlay networks for multimedia applications. Examples of overlay networks include application-layer multicast [85]–[88], Web content distribution networks, and

resilient overlay networks (RONs) [89]. Recently, there has been investigation on providing QoS support mechanism in overlay networks similar to the one in the Internet. OverQoS [90] aimed to provide architecture to offer QoS using overlay network. Service Overlay Networks [91] purchases bandwidth with certain QoS guarantees from individual network domains via bilateral service level agreement (SLA) to build a logical end-to-end service delivery infrastructure on top of existing data transport networks. Unlike the work on network-based QoS, research for QoS provisioning in application layer overlay has been pursued in an ad hoc manner. Thus, there is considerable room for improvement, especially in considering the video delivery requirement. Enabling video transport over ad hoc networks is another challenging task. The wireless links in an ad hoc network are highly error prone and can go down frequently because of node mobility, interference, channel fading, and the lack of infrastructure. In [92], Wang et al. proposed to combine multistream coding with multipath transport, to show that path diversity provides an effective way to combat transmission error in ad hoc networks. QoS routing [93] and QoS aware MAC [94] are two types of approaches to provide QoS for ad hoc networks from networking point of view. Extending the cross-layer framework to exploit the video delivery over ad hoc networks is also a quite interesting research direction. ACKNOWLEDGMENT The authors would like to thank Dr. B. Li from the Hong Kong University of Science and Technology for proofreading this manuscript. REFERENCES [1] D. Wu, Y. T. Hou, and Y.-Q. Zhang, “Transporting real-time video over the Internet: Challenges and approaches,” Proc. IEEE, vol. 88, no. 12, pp. 1855–1877, Dec. 2000. [2] P.-Y. Hébert, “End-to-end QoS in the user’s point of view,” in ITU Workshop on End-to-End Quality of Service. What Is It? How do we Get It?, Geneva, Switzerland, Oct. 1–3, 2003. [3] J. Liebeherr, “A framework for analyzing networks with deterministic and statistical QoS,” in Comet Group Seminar, Columbia University, New York, 2000. [4] J. Wroclawski, “The use of RSVP with IETF integrated services,” RFC 2210, Sep. 1997. [5] D. Grossman, “New terminology and clarifications for DiffServ,” RFC 3260, Apr. 2002. [6] Quality of Service (QoS) concept and architecture, 3GPP TS 23.107, Sept. 2003. [7] W. Li, “Overview of fine granularity scalability in MPEG-4 video standard,” IEEE Trans. Circuits Syst. Video Technol., vol. 11, no. 3, pp. 301–317, Mar. 2001. [8] F. Wu, S. Li, and Y.-Q. Zhang, “A framework for efficient progressive fine granularity scalable video coding,” IEEE Trans. Circuits Syst. Video Technol., vol. 11, no. 3, pp. 332–344, Mar. 2001. [9] A. Sampath, P. S. Kumar, and J. M. Holtzman, “Power control and resource management for a multimedia CDMA wireless system,” in Proc. IEEE PIMRC, vol. 1, 1995, pp. 21–25. [10] R. Rejaie, M. Handley, and D. Estrin, “Quality adaptation for congestion controlled video playback over the Internet,” in ACM SIGCOMM, 1999, pp. 189–200. [11] L. Qian, D. L. Jones, K. Ramchandran, and S. Appadwedula, “A general joint source-channel matching method for wireless video transmission,” in Proc. IEEE DCC, 1999, pp. 414–423.

ZHANG et al.: END-TO-END QoS FOR VIDEO DELIVERY OVER WIRELESS INTERNET

131

[12] C. Dovrolis, D. Stiliadis, and P. Ramanathan, “Proportional differentiated services: Delay differentiation and packet scheduling,” IEEE/ACM Trans. Networking, vol. 10, no. 1, pp. 12–26, Feb. 2002. [13] J. Shin, J. Kim, and C.-C. Jay Kuo, “Quality-of-service mapping mechanism for packet video in differentiated services network,” IEEE Trans. Multimedia, vol. 3, no. 2, pp. 219–231, Jun. 2001. [14] H. Zhang, “Service disciplines for guaranteed performance service in packet switching networks,” Proc. IEEE, vol. 83, no. 10, pp. 1374–1396, Oct. 1995. [15] Z.-L. Zhang, Z. Duan, and Y. T. Hou, “Virtual time reference system: A unifying scheduling framework for scalable support of guaranteed services,” IEEE J. Select. Areas Commun., vol. 18, no. 12, pp. 2684–2695, Dec. 2000. [16] V. Misra, W. Gong, and D. Towsley, “Fluid-based analysis of a network of AQM routers supporting TCP flows with an application to RED,” in Proc. ACM SIGCOMM, 2000, pp. 151–160. [17] S. Kunniyur and R. Srikant, “Analysis and design of an adaptive virtual queue algorithm for active queue management,” in Proc. ACM SIGCOMM, San Diego, CA, 2001. [18] J.-Y. Le Boudec, “Application of network calculus to guaranteed service networks,” IEEE Trans. Inform. Theory, vol. 44, no. 3, pp. 1087–1096, May 1998. [19] C.-S. Chang and J. A. Thomas, “Effective bandwidth in high-speed digital networks,” IEEE J. Select. Areas Commun., vol. 13, no. 6, pp. 1091–1100, Aug. 1995. [20] V. Firoiu, J.-Y. Boudec, D. Towsley, and Z.-L. Zhang, “Theories and models for Internet quality of service,” Proc. IEEE, vol. 90, no. 9, pp. 1565–1591, Sept. 2002. [21] D. Kitchin, “The 802.11 MAC protocol & quality of service,” lecture, Intel Corp.. [22] T. S. Rappaport, Wireless Communications: Principles & Practice. Englewood Cliffs, NJ: Prentice-Hall, 1996. [23] Q. Zhang and S. A. Kassam, “Finite-state Markov model for Rayleigh fading channels,” IEEE Trans. Commun., vol. 47, no. 11, pp. 1688–1692, Nov. 1999. [24] M. Zorzi, R. R. Rao, and L. B. Milstein, “Error statistics in data transmission over fading channels,” IEEE Trans. Commun., vol. 46, no. 11, pp. 1468–1477, Nov. 1998. [25] D. Wu and R. Negi, “Effective capacity: A wireless link model for support of quality of service,” IEEE Trans.Wireless Commun., to be published. [26] W. Kumwilaisak, Y. Hou, Q. Zhang, W. Zhu, C.-C. Kuo, and Y.-Q. Zhang, “A cross-layer quality of service mapping architecture for video delivery in wireless networks,” IEEE J. Select Areas Commun., vol. 21, no. 10, pp. 1685–1698, Dec. 2003. [27] A. Sehgal and P. A. Chou, “Cost-distortion optimized streaming media over DiffServ networks,” Proc. IEEE ICME, pp. 857–860, Aug. 2002. [28] W. Tan and A. Zhakor, “Packet classification schemes for streaming MPEG video over delay and loss differentiated networks,” in Proc. IEEE Packet Video Workshop, Kyongju, Korea, Apr. 2001. [29] J. Liu, B. Li, H.-R. Shao, W. Zhu, and Y.-Q. Zhang, “A proxy-assisted adaptation framework for object video multicasting,” IEEE Trans. Circuits Syst. Video Technol., to be published. [30] S. D. Servetto, K. Ramchandran, K. Nahrstedt, and A. Ortega, “Optimal segmentation of a VBR source for its parallel transmission over multiple ATM connections,” in Proc. IEEE Int. Conf. Image Processing, Santa Barbara, CA, Oct. 1997, pp. 5–8. [31] J. C. de Martin, “Source-driven packet marking for speech transmission over differentiated-service networks,” in Proc. IEEE Int. Conf. Acoustics, Speech, and Signal Processing, May 2001, pp. 753–756. [32] L. Xiao, M. Johansson, H. Hindi, S. Boyd, and A. Goldsmith, “Joint optimization of communication rates and linear systems,” submitted for publication. [33] M. Mirhakkak, N. Schult, and D. Thomson, “Dynamic bandwidth management and adaptive applications for a variable bandwidth wireless environment,” IEEE J. Select. Areas Commun., vol. 19, no. 10, pp. 1984–1997, Oct. 2001. [34] H. Schulzrinne, S. Casner, R. Frederick, and V. Jacobson, “RTP: A transport protocol for real-time applications,” RFC 3550, July 2003. [35] M. Handley and V. Jacobson, “SDP: Session description protocol,” RFC 2327, Apr. 1998. [36] H. Schulzrinne, A. Rao, and R. Lanphier, “Real time streaming protocol (RTSP),” RFC 2326, Apr. 1998. [37] S. Fu and M. Atiguzzaman, “SCTP: State of the art in research, products, and technical challenges,” IEEE Commun. Mag., vol. 42, no. 4, pp. 64–76, Apr. 2004.

132

[38] M. Handley, H. Schulzrinne, E. Schooler, and J. Rosenberg, “SIP: Session initiation protocol,” RFC 2543, 1999. [39] S. Jacobs and A. Eleftheriadis, “Streaming video using TCP flow control and dynamic rate shaping,” J. Visual Commun. Image Rep., vol. 9, no. 3, pp. 211–222, Sep. 1998. [40] W. Tan and A. Zakhor, “Real-time Internet video using error resilient scalable compression and TCP-friendly transport protocol,” IEEE Trans. Multimedia, vol. 1, pp. 172–186, Jun. 1999. [41] D. Disalem and H. Schulzrinne, “The loss-delay based adjustment algorithm: A TCP-friendly adaptation scheme,” in Workshop NOSSDAV, July 1998. [42] J. Padhye, V. Firoiu, D. Towsley, and J. Kurose, “Modeling TCP throughput: A simple model and its empirical validation,” in Proc. ACM SIGCOMM, Aug. 1998, pp. 303–314. [43] Equation based congestion control for unicast applications, S. Floyd, M. Handley, J. Padhye, and J. Widmer. [Online]. Available: http://www.aciri.org/tfrc [44] J. Widmer, R. Denda, and M. Mauve, “A survey on TCP-friendly congestion control,” IEEE Network, vol. 15, no. 3, pp. 28–37, May/Jun. 2001. [45] G. Montenegro, S. Dawkins, M. Kojo, V. Magret, and N. Vaidya, “Long thin networks,” RFC 2757, Jan. 2000. [46] A. Bakre and B. Badrinath, “I-TCP: Indirect TCP for mobile hosts,” in Proc. 15th Int. Conf. Distributed Computing Systems (ICDCS), Vancouver, BC, Canada, May 30–Jun. 2, 1995, pp. 136–143. [47] G. Cheung and T. Yoshimura, “Streaming agent: A network proxy for media streaming in 3G wireless networks,” in IEEE Packet Video Workshop, Pittsburgh, PA, Apr. 2002. [48] S. Cen, P. Cosman, and G. Voelker, “End-to-end differentiation of congestion and wireless loss,” in Proc. ACM Multimedia Computing and Networking, San Jose, CA, Jan. 2002, pp. 1–15. [49] D. Barman and I. Matta, “Effectiveness of loss labeling in improving TCP performance in wired/wireless network,” in Proc. 10th IEEE ICNP, Paris, France, Nov. 2002, pp. 2–11. [50] S. Biaz and N. Vaidya, “Discriminating congestion losses from wireless losses using inter-arrival times at the receiver,” in Proc. IEEE Symp. Application-Specific Systems and Software Engineering and Technology, Richardson, TX, Mar. 1999, pp. 10–17. [51] F. Yang, Q. Zhang, W. Zhu, and Y.-Q. Zhang, “End-to-end TCPfriendly streaming protocol and bit allocation for scalable video over wireless Internet,” IEEE J. Select Areas Commun., to be published. [52] Q. Zhang, W. Zhu, and Y.-Q. Zhang, “Resource allocation for multimedia streaming over the Internet,” IEEE Trans. Multimedia, vol. 3, no. 3, pp. 339–355, Sep. 2001. [53] K. Lai and M. Baker, “Nettimer: A tool for measuring Bottleneck link bandwidth,” in Proc. USENIX Symp. Internet Technologies and Systems, Mar. 2001. [54] M. Gerla, B. K. F. Ng, M. Y. Sanadidi, M. Valla, and R. Wang, “TCP Westwood with adaptive bandwidth estimation to improve efficiency/friendliness tradeoffs,” Comput. Commun. J., to be published. [55] N. Aboobaker, D. Chanady, M. Gerla, and M. Y. Sansadidi, “Streaming media congestion control using bandwidth estimation,” in Proc. IFIP/IEEE Int. Conf. Management of Multimedia Networks and Services, 2002. [56] R. Wang, M. Valla, M. Y. Sanadidi, and M. Gerla, “Using adaptive rate estimation to provide enhanced and robust transport over heterogeneous networks,” in Proc. 10th IEEE ICNP, Paris, France, Nov. 12–15, 2002. [57] Q. Zhang and S. A. Kassam, “Hybrid ARQ with selective combining for fading channels,” IEEE J. Select. Areas Commun., vol. 17, no. 5, pp. 867–880, May 1999. [58] R. Puri, K. Ramchandran, and A. Ortega, “Joint source channel coding with hybrid ARQ/FEC for robust video transmission,” in IEEE Multimedia Signal Processing Workshop, Redondo Beach, CA, Dec. 1998. [59] D. Wu, Y. T. Hou, and Y.-Q. Zhang, “Scalable video coding and transport over broad-band wireless networks,” Proc. IEEE, vol. 89, no. 1, pp. 6–20, Jan. 2001. [60] Q. Zhang, W. Zhu, and Y.-Q. Zhang, “Channel-adaptive resource allocation for scalable video transmission over 3G wireless network,” IEEE Trans. Circuits Syst. Video Technol., to be published. [61] B. Girod and N. Farber, “Wireless video,” in Compressed Video Over Networks. New York: Marcel Dekker, 2001. [62] Y. Shan and A. Zakhor, “Cross layer techniques for adaptive video streaming over wireless networks,” Proc. IEEE ICME, Aug. 2002.

PROCEEDINGS OF THE IEEE, VOL. 93, NO. 1, JANUARY 2005

[63] G. Cote, F. Kossentini, and S. Wenger, “Error resilience coding,” in Compressed Video Over Networks. New York: Marcel Dekker, 2001. [64] Y. Wang and Q.-F. Zhu, “Error control and concealment for video communications: A review,” Proc. IEEE, vol. 86, no. 5, pp. 974–997, May 1998. [65] A. Majumdar, D. Sachs, I. Kozintsev, K. Ramchandran, and M. Yeung, “Multicast and unicast real-time video streaming over wireless LANs,” IEEE Trans. Circuits Syst. Video Technol., vol. 12, no. 6, pp. 524–534, Jun. 2002. [66] T. Zhang and Y. Xu, “Unequal packet loss protection for layered video transmission,” IEEE Trans. Broadcast., vol. 45, no. 2, pp. 243–252, Jun. 1999. [67] G. Cheung and A. Zakhor, “Bit allocation for joint-source channel coding of scalable video,” IEEE Trans. Image Process., vol. 9, no. 3, pp. 340–356, Mar. 2000. [68] U. Horn, B. Girod, and B. Belzer, “Scalable video coding for multimedia applications and robust transmission over wireless channels,” in 7th Workshop Packet Video, Brisbane, Queensland, Australia, Mar. 1996, pp. 43–48. [69] J. Hagenauer, T. Stockhammer, C. Weiss, and A. Donner, “Progressive source coding combined with regressive channel coding for varying channels,” in Proc. 3rd ITG Conf. Source and Channel Coding, Munich, Germany, Jan. 2000, pp. 123–130. [70] G. Wang, Q. Zhang, W. Zhu, and Y.-Q. Zhang, “Channel-adaptive unequal error protection for scalable video transmission over wireless channel,” in Proc. SPIE VCIP, San Jose, CA, Jan. 2001, pp. 648–655. [71] M. van der Schaar and H. Radha, “Unequal packet loss resilience for fine-granular-scalability video,” IEEE Trans. Multimedia, vol. 3, no. 4, pp. 381–394, Dec. 2001. [72] S. Krishnamachari1, M. Schaar, S. Choi, and X. Xu, “Video streaming over wireless LANs: A cross-layer approach,” in Packet Video Workshop, 2003. [73] X. Xu, M. Schaar, S. Krishnamachari, S. Choi, and Y. Wang, “Adaptive error control for fine-granular-scalability video coding over IEEE 802.11 wireless LANS,” Proc. IEEE ICME, 2003. [74] M. Goel, S. Appadwedula, N. R. Shanbhag, K. Ramchandran, and D. L. Jones, “A low-power multimedia communication system for indoor wireless applications,” in Proc. IEEE Workshop SiPS’99, 1999, pp. 473–482. [75] Q. Zhang, W. Zhu, Z. Ji, and Y.-Q. Zhang, “A power-optimized joint source and channel coding for scalable video streaming over wireless channels,” in Proc. IEEE ISCAS, vol. 5, Sydney, NSW, Australia, May 2001, pp. 137–140. [76] M. Soleimanipour, W. Zhuang, and G. H. Freeman, “Modeling and resource allocation in wireless multimedia CDMA systems,” Proc. IEEE VTC, vol. 2, pp. 1279–1283, 1998. [77] S. L. Kim, Z. Rosberg, and J. Zander, “Combined power control and transmission rate selection in cellular networks,” Proc. IEEE VTC, vol. 3, pp. 1653–1657, 1999. [78] Y. Eisenberg, C. E. Luna, T. N. Pappas, R. Berry, and A. K. Katsaggelos, “Joint source coding and transmission power management for energy efficient wireless video communications,” IEEE Trans. Circuits Syst. Video Technol., vol. 12, no. 6, pp. 411–424, Jun. 2002. [79] Q. Zhang, Z. Ji, W. Zhu, and Y.-Q. Zhang, “Power-minimized bit allocation for video communication over wireless channels,” IEEE Trans. Circuits Syst. Video Technol., vol. 12, no. 6, pp. 398–410, Jun. 2002. [80] W. Jiang and H. Schulzrinne, “Modeling of packet loss and delay and their effect on real-time multimedia service quality,” in Proc. 10th Int. Workshop NOSSDAV, Chapel Hill, NC, Jun. 2000. [81] J. Bolot, T. Turleeti, and I. Wakeman, “Scalable feedback control for multicast video distribution in the Internet,” in Proc. ACM SIGCOMM, 1994, pp. 58–67. [82] V. Jacobson, S. Mccanne, and M. Vetterli, “Receiver-driven layered multicast,” in Proc. ACM SIGCOMM, Stanford, CA, Aug. 1996, pp. 117–130. [83] T.-W. A. Lee, S.-H. G. Chan, Q. Zhang, W. Zhu, and Y.-Q. Zhang, “Allocation of layer bandwidths and FEC’s for video multicast over wired and wireless networks,” IEEE Trans. Circuits Syst. Video Technol., vol. 12, no. 12, pp. 1059–1070, Dec. 2002. [84] J.-C. Liu, B. Li, and Y.-Q. Zhang, “An end-to-end adaptation protocol for layered video multicast using optimal rate allocation,” IEEE Trans. Multimedia, vol. 7, no. 6, pp. 87–102, Feb. 2004.

[85] S. Ratnasamy, M. Handley, R. Karp, and S. Shenker, “Application-level multicast using content-addressable networks,” in 3rd Int. Workshop on Networked Group Communication (NGC’01), London, U.K., 2001. [86] D. Pendarakis, S. Shi, D. Verma, and M. Waldvogel, “Almi: An application level multicast infrastructure,” in 3rd USENIX Symp. Internet Technologies and Systems (USITS), 2001, pp. 49–60. [87] M. Castro, M. B. Jones, A.-M. Kermarrec, A. Rowstron, M. Theimer, H. Wang, and A. Wolman, “An evaluation of scalable applicationlevel multicast built using peer-to-peer overlays,” in Proc. IEEE Infocom., San Francisco, CA, Apr. 2003. [88] V. N. Padmanabhan, H. J. Wang, and P. A. Chou, “Supporting heterogeneity and congestion control in peer-to-peer multicast streaming,” in 3rd Workshop for IPTPS, San Diego, CA, 2003. [89] D. Andersen, H. Balakrishnan, M. Kaashoek, and R. Morris, “Resilient overlay networks,” in Proc. ACM SOSP, 2001. [90] L. Subramanian, I. Stoica, H. Balakrishnan, and R. H. Katz, “OverQoS: Offering QoS using overlays,” in 1st Workshop on Hot Topics in Networks (HotNets-I), 2002. [91] Z. Duan, Z.-L. Zhang, and Y. T. Hou, “Service overlay networks: SLA, QoS, and bandwidth provisioning,” in Proc. ICNP, 2002. [92] S. Mao, S. Lin, S. S. Panwar, Y. Wang, and E. Celebi, “Video transport over ad hoc networks: Multistream coding with multipath transport,” IEEE J. Select. Areas Commun., vol. 21, no. 10, pp. 1721–1737, Dec. 2003. [93] C. Lin and J. Liu, “QoS routing in ad hoc wireless networks,” IEEE J. Select. Areas Commun., vol. 17, no. 8, pp. 1426–1438, Aug. 1999. [94] S. Kumar, V. S. Raghavan, and J. Deng, “QoS-aware MAC protocols for ad-hoc wireless networks: A survey,” Elsevier Ad-Hoc Network J., to be published. [95] M. van der Schaar and H. Radha, “Adaptive motion compensation fine-granular-scalability (AMC-FGS) for wireless video,” IEEE Trans. Circuits Syst. Video Technol., vol. 12, no. 6, pp. 360–371, Jun. 2002.

Qian Zhang (Senior Member, IEEE) received the B.S., M.S., and Ph.D. degrees from Wuhan University, Wuhan, China, in 1994, 1996, and 1999, respectively, all in computer science. She joined Microsoft Research, Asia, Beijing, China, in July 1999. Now, she is the research manager of the Wireless and Networking Group. She has published more than 80 refereed papers in international leading journals and key conferences in the areas of wireless/Internet multimedia networking, wireless communications and networking, and overlay networking. She is the inventor of about 20 pending patents. Her current research interest includes seamless roaming across different wireless networks, multimedia delivery over wireless, Internet, next-generation wireless networks, and P2P network/ad hoc network. Dr. Zhang is a member of the Visual Signal Processing and Communication Technical Committee and the Multimedia System and Application Technical Committee of the IEEE Circuits and Systems Society. She is also a Member and Chair of QoSIG of the Multimedia Communication Technical Committee of the IEEE Communications Society. Dr. Zhang is now serving as Associate Editor of IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY. She is also serving as Guest Editor for a special issue on wireless video in IEEE Wireless Communication Magazine. Dr. Zhang has recently received the TR 100 (MIT Technology Review) World’s Top Young Innovator Award.

ZHANG et al.: END-TO-END QoS FOR VIDEO DELIVERY OVER WIRELESS INTERNET

133

Wenwu Zhu (Senior Member, IEEE) received the B.E. and M.E. degrees from the National University of Science and Technology, Changsha, China, in 1985 and 1988, respectively, the M.S. degree from Illinois Institute of Technology, Chicago, in 1993, and the Ph.D. degree from Polytechnic University, Brooklyn, NY, in 1996, all in electrical engineering. From August 1988 to December 1990, he was with the Graduate School, University of Science and Technology of China (USTC), and Chinese Academy of Sciences (Institute of Electronics), Beijing, China. He joined Microsoft Research, Beijing, in 1999 as a Researcher in the Internet Media Group, and now is Research Manager of Wireless and Networking Group. Prior to his current post, he was with Bell Labs., Lucent Technologies, Murray Hill, NJ, as a Member of Technical Staff during 1996–1999. He has published over 160 refereed papers in various key journals and conferences in the areas of wireless/Internet multimedia delivery, wireless communications and networking, and has contributed to the IETF ROHC WG draft on robust TCP/IP header compression over wireless links. He is inventor of more than a dozen pending patents. His current research interest is in the area of wireless/Internet multimedia communication and networking, and wireless communication and networking. Dr. Zhu served as Guest Editor for the special issues on “Streaming Video” and special issue on “Wireless Video” in IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY. He also served as a Guest Editor for the special issue on “Advanced Mobility Management and QoS Protocols for Wireless Internet” in IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS. Currently, he is serving as a Guest Editor for the special issue on “Advanced Video Coding and Delivery” in PROCEEDINGS OF THE IEEE, and a Guest Editor for the special issue on “Wireless Video” in IEEE Wireless Communication Magazine. Currently he is Associate Editor for IEEE TRANSACTIONS ON MOBILE COMPUTING, IEEE TRANSACTIONS ON MULTIMEDIA, and IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, respectively. He received the Best Paper Award in IEEE Transactions on Circuits and Systems for Video Technology in 2001. He is also the Chairman of IEEE Circuits and System Society Beijing Chapter and the Secretary of Visual Signal Processing and Communication Technical Committee. He is a member of Eta Kappa Nu, Multimedia System and Application Technical Committee and Life Science Committee in IEEE Circuits and Systems Society, and Multimedia Communication Technical Committee in IEEE Communications Society.

134

Ya-Qin Zhang (Fellow, IEEE) received the B.S. and M.S. degrees in electrical engineering from the University of Science and Technology of China (USTC), Hefei, Anhui, China, in 1983 and 1985, respectively, and the Ph.D. degree in electrical engineering from George Washington University, Washington, DC, in 1989. He is currently the Corporate Vice Present of the Mobile and Device Group at Microsoft Corporation, Redmond, WA. He is responsible for product development of Microsoft’s Mobile and Embedded Division, including the WinCE operating system, Smartphone, PocketPC, and other Windows Mobile platform and devices. Prior to that, he was the Managing Director of Microsoft Research Asia from 1999 to 2004. Previously, he was the Director of the Multimedia Technology Laboratory, Sarnoff Corporation, Princeton, NJ (formerly David Sarnoff Research Center and RCA Laboratories). Prior to that, he was with GTE Laboratories Inc., Waltham, MA, from 1989 to 1994. He has been engaged in research and commercialization of MPEG2/DTV, MPEG4/VLBR, and multimedia information technologies. He has authored and co-authored over 200 refereed papers in leading international conferences and journals, and has been granted over 40 U.S. patents in digital video, Internet, multimedia, wireless, and satellite communications. Many of the technologies he and his team developed have become the basis for start-up ventures, commercial products, and international standards. He serves on the Board of Directors of five high-tech IT companies and has been a key contributor to the ISO/MPEG and ITU standardization efforts in digital video and multimedia. Dr. Zhang served as the Editor-In-Chief for the IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY from July 1997 to July 1999. He was the Chairman of the Visual Signal Processing and Communications Technical Committee of the IEEE Circuits and Systems (CAS) Society. He serves on the editorial boards of seven other professional journals and over a dozen conference committees. He has received numerous awards, including several industry technical achievement awards and IEEE awards, such as the CAS Jubilee Golden Medal. He was named “Research Engineer of the Year” in 1998 by the Central Jersey Engineering Council for his “leadership and invention in communications technology, which has enabled dramatic advances in digital video compression and manipulation for broadcast and interactive television and networking applications.” He recently received The Outstanding Young Electrical Engineer of 1998 award.

PROCEEDINGS OF THE IEEE, VOL. 93, NO. 1, JANUARY 2005