M. W. Garrett and W. Willinger. Analysis, modeling and generation of self-similar VBR video trac. In Proceedings of SIGCOMM, August 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....or an UltimateTV [19] STB. A very large buffer would also allow a more efficient utilization of the channel bandwidth. Future VOD services will distribute videos in compressed form and the bandwidth requirements of these videos will depend on the rate at which the images being displayed change [1, 8]. For instance, daytime action scenes and cartoons will require more bandwidth than slower moving scenes and night scenes. The buffer sizes of conventional diskless STBs are limited by memory cost considerations. Hence buffer overflow is as a serious concern as buffer underflow. As a result, the ....

Garrett, M. and W. Willinger. Analysis, modeling and generation of self-similar VBR video traffic. Proceedings of the ACM SIGCOMM '94 Conference, pages 269--280, Aug. 1994.

....tra#c model is introduced and used to estimate the queue size of the queuing system associated to that tra#c model. Key words. queuing systems, overflow probability, Fractional Brownian Motion. AMS Subject Classification. 60K25 1. Introduction Over the last years many studies ( 1] 3] [7], 10] 11] 13] 16] 19] 20] have shown that packet cell tra#c through telecommunication networks (like Ethernet, LAN, WAN, ISDN, ATM) exhibits long range dependence and selfsimilarity, i.e. the autocorrelation function decays asymptotically as a power function with negative exponent, and ....

M. W. Garrett, W. Willinger, Analysis, Modeling and Generation of SelfSimilar VBR Video Tra#c, Proceedings of ACM SIGComm '94, London, pages 269-280, 1994.

.... service with stringent requirements is a bounded delay service, which guarantees that all traffic from a flow satisfies worstcase delay bounds and that no packets are dropped [4] An example of bursty traffic is MPEG 1 video traffic, which has correlations of traffic over multiple time scales [5]. The burstiness of MPEG traffic is illustrated in Figure 1 for a sequence from an MPEG 1 compressed motion picture. In [13] it was shown that packet networks with a bounded delay service for MPEG 1 video traffic can capture multiplexing gain. There are several approaches to improve multiplexing ....

M. W. Garrett and W. Willinger. Analysis, modeling and generation of self-similar vbr video traffic. In Proc. ACM Sigcomm `94, pages 269--280, August 1994.

....fractality and long range dependency, in network traffic. The seminal paper by Leland et al. 28] showed self similarity in Ethernet traffic. Later, many others revealed similar properties in wide area traffic, signaling traffic, high speed network traffic and in multimedia and video traffic [6, 9, 10, 11, 15, 17, 18, 32]. Many studies have shown that ignoring the self similarity in the analysis of systems leads in general to an underestimation of important performance measures [32] Additionally, various studies have shown that the presence of self similarity is generally associated with the presence of ....

....function. From (2) we obtain (for details, see [27, Section 1.4.1. 2] 46587 [ fe 9 CEg ihj f Dkl Q b cmh cn;VK (3) There are various methods to identify self similar processes, such as R S analysis, Whittle s maximum likelihood estimator, or the variance time plot method [18, 28]. For the purpose of this paper, it suffices to use the variance time plot method. This method estimates the Hurst parameter from a graph of var 4L5M7 versus , plotted on a log log scale. An example of such a variance time plot is given in Figure 2, which will be discussed later. From ....

M. W. Garrett and W. Willinger. Analysis, modeling and generation of self-similar VBR video traffic. In Proc. ACM SIGCOMM '94, volume 24 of Computer Communications Review, pages 269--280, London, October 1994.

....etc. However, compressed video, especially variable bit rate (VBR) video, typically exhibits significant burstiness on multiple time scales, owing to the encoding schemes and the content variation # A preliminary version of this paper appeared at SIROCCO 2001. between and within video scenes [6], 9] This burstiness complicates the design of e#cient transport mechanisms for such media. In a network where resources such as bandwidth and bu#ering capacity are constrained there is a need for an e#cient video delivery system that can achieve high resource utilization and maximize the QoS ....

Garrett, M.---Willinger, W.: Analysis, Modeling and Generation of Self-Similar VBR Video Tra#c. Proceedings of ACM SIGCOMM, pp. 269--280, September 1994.

....of compressed video and audio data. The compression causes a variable data rate of each TV program because scenes with a lot of motion in the picture are encoded with a higher bit rate than scenes with less motion. This bursty characteristic of variable bit rate (VBR) compressed video traffic [9, 10] requires allocation of extra bandwidth to achieve delay and bit error rate (BER) requirements of DVB. In order to get a fixed data rate for the transmission, stuffing packets are inserted into the TS [7] Within MPEG 2 TS, it is also possible to carry defined data containers in addition to ....

....packets. Therefore the case where Internet traffic is carried on the unused bandwidth left from video traffic is analyzed in this thesis. #.5. Problem Statement An important feature of video traffic at the packet level that has a significant impact on performance is traffic correlation [9, 10]. Specifically, bursty traffic patterns generated by VBR compressed video and audio tend to exhibit certain degrees of correlation between arrivals, and show long range dependence in time (self similar traffic) Studies of Leland, Taqqu, Willinger, Crovella, Beran and Paxon [11, 12, 13] have ....

[Article contains additional citation context not shown here]

Garrett, M. W. and W. Willinger, Analysis, modeling and generation of self-similar VBR video traffic, Proceedings of the ACM Sigcomm '94, pp. 269-280, 1994.

.... studies have also observed the same self similar nature in other types of traffic as in wide area networks (WANs) 27, 19] in the SS7 data of the Common Channel Signaling Networks (CCSN) 7] in the traffic due to the World Wide Web (WWW) 6] or even in Variable Bit Rate (VBR) video traffic [12, 4]. All these findings have lead to new insights into the stochastic modeling of the traffic processes and into the performance evaluation of high speed networks. However, recently Eramilli et al. 9] pointed out the relative shortage of results in network analysis with self similar traffic, ....

M. W. Garrett and W. Willinger. Analysis, modeling and generation of self-similar VBR video traffic. In Proc. ACM SIGCOMM '94, pages 269--280, London, UK, 1994.

....end user perceived rate fluctuations depend on the end to end path properties. The phenomenon of self similarity was observed in data networks in [14] 15] followed by several experimental papers showing fractal characteristics in other types of networks and traffic, e.g. in video traffic [3] [11] or in ATM networks [18] A comprehensive bibliographical guide is presented in [33] These observations have questioned the validity of previous short memory models when applied to network performance analysis [25] The impact of self similar models, e.g. Fractional Gauss Noise on queuing ....

M. Garrett and W. Willinger. Analysis, modeling and generation of self-similar VBR video traffic. In Proc. ACM SIGCOMM, pages 269--280, 1994.

.... in a packet based network such as the Internet, models of the traffic once assumed to be Markovian are now also best represented by heavy tail distributions [Paxson95, Crovella97, Feldmann98b] Examples of VBR traffic that may be characterised by heavy tail distributions include video ( Garrett94] Local Area Network (LAN) traffic ( Leland94] and WWW transactions carried over TCP IP (e.g. Crovella97] The reason that the heavy tail distribution more accurately characterises Internet traffic (or the duration of telephone calls) is because each of these sources, when multiplexed ....

....that this behaviour is best represented as a multifractal process. Feldmann98a] then showed how cascades of (fractal) processes can be used to recreate the behaviour of WAN and LAN traffic. Self similarity is not a property unique to Internet traffic, a number of previous studies, notably 35 [Garrett94] reported the presence of self similarity in VBR video traffic. Interestingly, Ryu96] reported in studies of VBR video traffic carried via ATM, that the loss rates are not adversely affected by the self similarity properties and that Markovian models are sufficient for performance analysis. This ....

[Article contains additional citation context not shown here]

M.E. Garrett and W. Willinger. Analysis, Modeling and Generation of SelfSimilar VBR Video Traffic. In Proceedings of ACM SIGCOMM'94, London, UK, August 1994. (pp 33, 36, 44)

....second simulation topology, where the upper two branch links have limited bandwidth, while the link on the third branch is congested by high background traffic load. This topology is not as symmetric as the first one. Background traffic is injected using a real traffic trace (the Star Wars movie [11]) One Star Wars flow is transmitted on each of the three main branches starting from router r2 to a randomly chosen receiver on each branch, so as not to create a bottleneck on the main shared path. In both topologies, three groups of flows comprise the expected partitioning: one group for each ....

M. W. Garrett and W. Willinger. Analysis, Modeling and Generation of Self-Similar VBR Video Traffic. In Proceedings of ACM SIGCOMM Conference, pages 269--280, London, UK, 31st - 2nd 1994.

....varies through a 10 minute extract from three videos: Starwars, a Bond movie and the Simpsons cartoon [18] when coded with a constant quantization parameter. It has, in fact, been observed that rate variations in movies persist even as the rate is averaged over increasingly long intervals [1,5]. This property is know mathematically as self similarity or long range dependence and can have serious consequences on the feasibility of multiplexing video sources with performance guarantees [14] One manifestation of long range dependence is that periods when the average rate is high can last ....

....that periods when the average rate is high can last for significant periods of time (see Figure 1) If an overload occurs in a network multiplexer, congestion can persist, notably with unfortunate consequences for video quality. The following is a mathematical definition of long range dependence [1,5]: 0 2 4 6 8 2 2.5 3 3.5 4 4.5 5 Constrained VBR Open Loop c (in Mbits s) b(c) in Mbits) up Mbits Figure 2. Burstiness function of Open Loop and Constrained VBR coding Let X i be the size of the i th frame and define the aggregate process X k = k m Gamma1 i=k X i : A ....

[Article contains additional citation context not shown here]

M.W. Garrett and W. Willinger. Analysis, Modeling and Generation of Self-Similar VBR Video Traffic. In Proceedings of SigComm. ACM, September 1994.

....with real network traffic, in particular traffic that exhibits fluctuations on a wide range of time scales. The simulation is based on a compressed video trace of the Star Wars movie, which has been extensively studied in the literature and been shown to exhibit fluctuations on all time scales [8] . In Figure 7 and 8, we plot the measured overflow probability p f as a function of the critical time scale, for different values of the high pass variance estimation window T s = K f T h . 1 10 1 T as a function of 1 T (Starwars, T =T =1.0e 3, n=10000) ....

M. W. Garrett and Walter Willinger. Analysis, Modeling and Generation of Self-Similar VBR Video Traffic. In Proc. ACM SIGCOMM '94, pages 269--280, London, UK, August 1994.

....in real time [5] namely taking into account the traffic process history up to time # solely. Thus, the predictor can be used not only for stored video but also for live Internet traffic. It has been shown [3] 5] that the linear predictor defined in 1 proves adequate for prediction of VBR video [7]. Most interestingly, even though VBR video shows long range dependence, only the short scale correlations determine the prediction error, with the long range correlations having an slight incremental value [3] Such findings are consistent with recent contributions in network performance ....

M.W. Garrett and W. Willinger. Analysis, modeling and generation of self-similar VBR video traffic. In Proceedings of the 1994.

No context found.

M. W. Garrett and W. Willinger. Analysis, modeling and generation of self-similar VBR video trac. In Proceedings of SIGCOMM, August 1994.

No context found.

M. W. Garrett and W. Willinger. Analysis, Modeling and Generation of Self-Similar VBR Video Traffic. In Proc. ACM Sigcomm `94, pages 269--280, August 1994.

No context found.

M. W. Garrett and W. Willinger. Analysis, modeling and generation of self-similar vbr video tra#c. In Proc. ACM SIGCOMM, pages 269--280, 1994.

No context found.

M. W. Garrett and W. Willinger. Analysis, Modeling and Generation of Self-Similar VBR Video Traffic. In Proceedings of ACM SIGCOMM Conference, pages 269--280, London, UK, 31st - 2nd 1994.

No context found.

M. Garret and W. Willinger. Analysis, modeling and generation of self-similar VBR video traffic. In SIGCOMM Symposium on Communications Architectures and Protocols, pages 269--280, London, UK, Sept. 1994.

No context found.

M. Garrett and W. Willinger, #Analysis, modeling and generation of self-similar VBR video tra#c," inProc. ACM SIGCOMM, pp. 269#280, August 1994.

No context found.

M. W. Garrett and Walter Willinger. Analysis, Modeling and Generation of Self-Similar VBR Video Trac. In Proc. ACM SIGCOMM '94, pages 269-280, London, UK, August 1994.

No context found.

M. Garrett and W. Willinger. Analysis, modeling and generation of self-similar VBR video traffic. In Proc. ACM SIGCOMM, pages 269--280, 1994.

No context found.

M. Garrett and W. Willinger, Analysis, modeling and generation of self-similar VBR video traffic, in: Proc. of ACM SIGCOMM'94, 1994, pp. 269--280.

No context found.

M. W. Garrett and W. Willinger. Analysis, modeling and generation of self-similar VBR video tra#c. In Proceedings of the ACM SIGCOMM '94 Conference, pages 269--280, 1994. 14

No context found.

M. Garrett and W. Willinger. Analysis, modeling and generation of self-similar VBR video traffic. Proceedings of ACM SIGCOMM, pp. 269-280, September 1994.

No context found.

M.W. Garrett and W. Willinger. Analysis, modeling and generation of self-similar VBR video tra#c. In Proc. ACM SIGCOMM, September 1994.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC