Quality of Service for Intenet Multimedia
|
Description
|
Today, the Internet carries increasingly diverse applications that require different levels ofnetwork quality of service (QoS). In particular, continuous media (CM) applications have muchmore stringent QoS requirements than traditional Transmission Control Protocol (TCP)-basedapplications. Be-cause we face a variety of application requirements, Internet QoS will remaina challenging task even if bandwidth becomes abundant and cheap. To provide an appropriate level of service quality, the current best-effort Internet must besupplemented with the ability to distinguish traffic from different service classes and providedifferent levels of service for these classes. The differentiated services (DiffServ, or DS)framework of the Internet Engineering Task Force (IETF) intends to provide appropriate gradesof service to applications having different requirements by using relatively simple yetscalable methods. The deployment of such service-differentiating networks, combined with theend-system application's clever use of service features, will greatly benefit multimediaapplications. There has been little research focused on the "QoS interaction" between CMapplications and QoS-enabled networks. The main focus of this book is to investigate positiveinteractions between multimedia applications and service-differentiating networks. Therefore,we will present a futuristic QoS mapping framework to combine application-level andnetwork-level efforts for QoS support. First, we will propose a video categorization scheme based on the concept of a relativepriority index (RPI).We define this RPI on the basis of a content segment's QoS needs.Subsequently, we pose the problem of mapping RPIs to the classes of service offered by aservice-differentiating network. We refer to this mapping as "QoS mapping." Next, we willpresent our design for effective QoS mapping from categorized application data segments to anetwork's service classes under the constraints imposed by either the system or cost. An RPIserves as a good bridging tool in enabling a network to be content-aware and thus to achievebetter end-to-end video quality. Finally, we will derive practical guidelines for effective QoSmapping. Second, we will present an adaptive packet-forwarding mechanism that can be employed by anetwork to enhance cooperation with an application. The main goal of this packet-forwardingmechanism is to maintain the in-tended measure of QoS discrimination among different DiffServclasses (i.e., grades) more persistently in the presence of network traffic fluctuation. Thispacket-forwarding mechanism is based on widely accepted queue management and scheduling schemesto be implemented with DiffServ. Third, we will present our design of a dynamic QoS mapping control scheme for DiffServnetworks. This scheme adjusts QoS mapping online and includes feedforward and feedback QoScontrol. This control scheme can be used by what we call a "CM gateway," which is placed at theedge of aDiffServdomain. The CM gateway dynamically performs QoS mapping for multipleaggregated CM flows. We have developed a dynamic and aggregate QoS mapping control scheme to beused by a CM gateway in order to achieve reliable and consistent end-to-end quality provisionfor individual CM streams with relative service differentiation of categorized packets. Inparticular, we have focused on dynamic QoS control to handle the QoS demand variations of CMapplications (e.g., ying priorities from aggregated/ categorized packets) and the QoS supplyvariations of a DiffServ network (e.g., variations in loss/delay due to fluctuating networkloads). Fourth, we will investigate a fine-grained scalable (FGS) version of video streaming based onthe MPEG-4 standard. This version prioritizes stream delivery over loss-rate differentiatingnetworks. Our proposed system is focused on seamless integration of rate adaptation andprioritized packeti- zation. Also,the proposed system simplifies differentiation for MPEG-4 FGSvideo streaming. The system consists of three key components: (1) rate adaptation with scalablesource encoding; (2) content-aware, prioritized packetization; and (3) loss-based differentialforwarding. Fifth, we will present a system that combines application-level unequal error protection (UEP)using forward error correction (FEC) with network- level QoS support using the resourcereservation protocol (RSVP)/integrated services or DiffServ. We have modeled this system, andwe will present our evaluation of its performance. Finally, we will present an enhanced network service model for layered video multicastapplications for DiffServ networks, which consists of enhanced active queue management (AQM)and hierarchical priority marking. We will present three sets of extended multiple random earlydetection (RED) parameters to enhance the AQM model for generic service differentiation andsupport of receiver-driven layer multicast (RLM) traffic. |
|
