A rate control mechanism for packet video in the Internet zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Jean-Chrysostome Bolot Thierry Turletti INRIA B. P. 93 06902 Sophia- Antipolis Cedex France {bolot, turletti}@sophia.inria.fr Datagram networks such zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA make it essentially impossible to provide the guaran- Abstract tees, generally expressed in terms of minimum band- as the Internet do not pro- width or maximum delay, required by the so-called vide guaranteed resources such as bandwidth or guar- real-time applications such as videoconferences [16]. anteed performance measures such as maximum delay. Two distinct approaches have emerged recently to One way to support packet video in these networks is tackle this problem. to use feedback mechanisms that adapt the output One approach is to extend current protocols and rate of video coders based on the state of the net- work. In this paper, we present one such mechanism. switch scheduling disciplines to provide the desired We describe the feedback information, and how it is performance guarantees. This approach requires that used by the coder control algorithm. We also examine admission control, policing, reservation, and/or so- how the need to operate in a multicast environment phisticated scheduling mechanisms be implemented in impacts the design of the control mechanism. the network. The design, analysis, and evaluation Our mechanism has been implemented in the H.261 of such mechanisms is an active research area (e.g. video coder of IVS. IVS is a videoconference system [3, 16, I S ] ) . These mechanisms will eventually be im- for the Internet developed at INRIA. Experiments in- plemented in the Internet, but they are not expectred dicate that the control mechanism is well suited to to be available in the very near future. An exam- the Internet environment. In particular, it makes it ple videoconference system based on this approach is possible to establish and maintain quality videocon- described in [5]. Another approach is to control the ferences even across congested connections in the In- ternet. Furthermore, it prevents video sources from rate at which packets can be sent over a connection, swamping the resources of the Internet, which could the objective being to limit this rate to the capacity of lead to unacceptable service to all users of the net- the connection. We observed earlier that this capacity work. changes with time. Therefore, the control mechanism must be somehow informed of such changes. One way is for sources of packets to receive feedback about the 1 Introduction state of the network and to control the rate at which packets are sent into the network accordingly. Data networks such as the Internet use datagram We propose to use this second approach to control switching as a means of dynamically allocating net- sources of real-time traffic. In this paper, we consider work resources on a demand basis. Datagram switch- specifically sources of video traffic, i.e. video coders. ing facilitates the interconnection of networks with The goal then is to use feedback information about $3.00 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA different architectures and provides flexible resource 1216 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA the state of the network to adapt the output rate of allocation and good reliability against node and link the coder. failure. However, datagram switching coupled with the FCFS discipline typically used in current switches This approach has two important advantages. The 9c.3.1 0743-166X/94 0 1994 IEEE
does not require special support from the network such zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA first advantage is that it is in fact surprisingly easy to The variable rate video stream is sent into a buffer feedback zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA which is drained at a constant rate. The amount of control the output rate of a software video coder by adjusting parameters inside the coder. Specific ways data in the buffer is used as a feedback information by to do this will be described in Section 2. a controller which adapts the parameters of the coder, The second advantage of this approach is that it and hence the output rate of the coder, in order to prevent buffer overflow or underflow (e.g. [2]). of multicast transmission zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA as admission control, resource allocation, etc. There- fore, it can easily be implemented in the current In- ternet. This is very important because the increased computing power of workstations and the availability Feedback mechanisms for video sources have zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA well as audio and video [6] as applications such as VAT [12], NV [8], NEVOT [23], W and IVS has led to a huge increase of real-time traffic in the Internet. For example, IETF meetings, con- Figure 1: Feedback control for networks with fixed ferences (e.g. the 4th European Networking Confer- capacity channels ence), and seminars (e.g the Xerox PARC Thursday seminars) are now regularly broadcast. The uncon- also trolled transmission of audio and video streams could been proposed for networks with variable capacity easily swamp the resources of the Internet and lead to channels such as the Internet. There, the goal is to unacceptable service for all users of the network. In- adjust the parameters (and hence the output rate) cluding control mechanisms in real-time applications of video coders based on feedback information about helps prevent such situations from happening. video zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA changing network conditions, i.e. changing capacity Feedback control mechanisms are already used in in the network (Figure 2). References [9, 251 propose the Internet to control sources of non real-time traf- fic. The best example is TCP, where the feedback feedback information is packet losses detected by timeouts or multiple acknowledgements at the source, and the con- trol scheme is Van Jacobson’s dynamic window scheme raw video coder An important goal of TCP and other mechanisms [ll]. for data traffic is to maximize throughput and mini- mize packet delay or loss, or equivalently to optimize some function of throughput and delay such as power Figure 2: Feedback control for networks with variable [15]. However, we note that this goal does not neces- capacity channels sarily translate into optimizing the quality of a video transmission. Therefore, we do not expect that the to use feedback control, but they do not describe spe- mechanisms proposed for the control of data traffic cific control mechanisms. Reference [26] describes a will be suitable for video traffic. specific scheme. However, this scheme requires that The idea of controlling sources of video traffic to the source of a connection know the service rate of adapt to the capacity available in a network is not the bottleneck on this connection. This service rate new. Video has conventionally been transmitted over can be estimated in networks where the switches use specific networks which provide connections with con- a so-called Fair Queueing or equivalent discipline [18]. stant or nearly constant capacity channels (e.g. tele- However, it is not available in networks with FCFS phone or CATV networks). However, the rate of a switches such as the Internet. Reference [14] describes video sequence can vary rapidly with time due to the a feedback control scheme which requires that switches effects of scene complexity and motion. The problem, send their buffer occupancies and service rates back to therefore, is to obtain from the variable rate sequence the source. Reference [13] describes a mechanism in a constant rate stream of data that can be sent into the which the time at which video packets are sent (and network. This is typically done as shown in Figure 1. hence the rate at which the image is refreshed at the 9c.3.2 1217
Recommend
More recommend