Jean-Chrysostome Bolot. Characterizing end-to-end packet delay and loss in the Internet. Journal of High Speed Networks, 2(3):305-323, 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....a session, the conditional CDF can be applied directly to estimate the burstiness of late losses. If an adaptive playout delay is used, we cannot directly relate the conditional CDF to late loss burstiness. We have found little known literature on the topic of conditional delay dependency. Bolot [1] analyzed the conditional property of round trip delays of consecutive packets. Their conclusion is that such delays have a random variation in lightly loaded conditions, and when background trac load is high, consecutive delays often exhibit spikes. A delay spike is a sequence of delays that ....

Jean-Chrysostome Bolot. Characterizing end-to-end packet delay and loss in the Internet. Journal of High Speed Networks, 2(3):305-323, 1993.

....performance applications is persistently much less than the end to end structural and load characteristics of the network would indicate is available. One source of poor TCP throughput that has been identified is a packet loss rate that is much greater than what would be reasonably expected [20]. Packet loss is interpreted by TCP as an indication of network congestion between a sender and receiver. Packet loss, however, may be due to random factors other than network congestion, such as intermittent hardware faults [4] Current efforts to improve end to end performance are taking ....

....at random intervals in bursts of multiple packets, rather than single packet drops. Borella[33] found bursty packet loss behavior as well. Additionally, the probability of a packet loss event increases when packets are queued in intermediate hops as the network becomes more loaded. Bolot[20] found that packet loss demonstrates random characteristics when the stream uses a fraction of the available network bandwidth. As the number of multiple TCP connections increases, the behavior of each packet loss factor p i is unaffected as long as few packets are queued in routers or ....

Jean-Chrysostome Bolot. "Characterizing Endto -End packet delay and loss in the Internet.", Journal of High Speed Networks, 2(3):305--323, 1993.

....INTRODUCTION Methods for discovering network internal characteristics using measurements taken at endpoints are increasingly valuable as applications and services seek to adapt to network properties. Previous work on estimation of network properties has focused on estimating bottleneck bandwidth [3], 5] 22] 17] and estimating per link bandwidth, latency, and loss This work was partially supported by NSF research grant ANI 9986397. 15] 19] 7] 23] While each of these techniques is useful, an important network resource that has not previously been addressed is the configuration of ....

....of two packets that queue back to back in front of the link being measured. This results in an effect (originally illustrated in [14] in which the inter packet gap of the two packets leaving the bottleneck is inversely proportional to the link bandwidth. This phenomenon was explored further in [3] and developed into a tool in [5] A number of further fruitful directions have followed, including the use of groups of packets and single packets [22] 17] 18] 12] While some of these methods use pairs of packets, just as the loss pair method does, the loss pair method focuses on relating ....

Jean-Chrysostome Bolot. Characterizing end-to-end packet delay and loss in the Internet. Journal of High Speed Networks, 2(3):305--323, 1993.

....algorithm, in the next section, makes adjustments it changes the value of mActive accordingly. This algorithm suffices for symmetric distributions of transit variations. In reality, outliers exist in the distribution of transit times, particularly those arising from packet compression events [1] that adversely influence the mapping offset. An additional mechanism, derived from Ramjee et al. s algorithm 4 [10] is therefore employed to identify periods of packet compression. During these periods the estimate of the mapping offset is not updated. Compression periods are typically of the ....

Jean-Chrysostome Bolot. Characterizing end-to-end packet delay and loss in the Internet. Journal of High Speed Networks, 2(3):305--323, 1993.

....f(t) P [d i jd i l 2 (t 1 ; t 2 ) 11) Similarly, we can have: f(t) P [d i jd i 1 ; d i 2 ; d i l 2 (t 1 ; t 2 ) 12) There are many kind of metrics and functions we can use to characterize conditional delay dependency. The author have found little known literature on this topic. Bolot [5] analyzed the conditional property of consecutive delays. We have examined both (9) and (11) For (11) the results are intuitive, for previous high delays, the conditional pdf also shifts toward right. Among all these metrics, it seems ccdf is more useful, especially in real time multimedia ....

Jean-Chrysostome Bolot. Characterizing end-to-end packet delay and loss in the Internet. Journal of High Speed Networks, 2(3):305-323, 1993.

....of service uniformity. Figure 2 shows two plots of measurements taken while OC was experiencing some network instability. The New Jersey Germany data might indicate route flapping between two, or more, different paths. It is possible that the results of Figure 2a could be due to congestion [Bolo93] either on the PVC, or on the carrier s network. Congestion on the PVC was unlikely in this case since the test packets were essentially the only traffic. Visual inspection of the plots in Figures 1 and 2 suggest that something has changed for the worse in Figure 2. Since the ultimate goal ....

Jean-Chrysostome Bolot, "Characterizing End-to-End Packet Delay and Loss in the Internet, " Journal of High Speed Networks, Volume 2, Number 3, pp 305-323, 1993.

....ALGORITHM In the section 3 we looked at an adaptive method of playout for audiovideo frames. The success of this method depends on the accuracy of estimating forecasting the value of the end to end delay of packets. Analysis of delay characteristics has been carried out by several researchers [Bolot, 1993] by collecting traces from the Internet. These traces show occurrence of frequent spikes which last for several milliseconds. A spike is characterised by a sudden large increase in the end to end delay of packets. The delay trace plotted in Figure 6 was collected between a host at the University ....

Jean-Chrysostome Bolot. Characterizing end-to-end packet delay and loss in the Internet. Journal of High Speed Networks, 2(3):305--323, 1993.

....as: #(#) # [# # ## ### # (# # ## # ) 11) Similarly,we can have: #(#) # [# # ## ### ## ### # #### # ### # (# # ## # ) 12) There are many kind of metrics and functions we can use to characterize conditional delay dependency. The author have found little known literature on this topic. Bolot [5] analyzed the conditional propertyof consecutive delays. Wehave examined both (9) and (11) For (11) the results are intuitive, for previous high delays, the conditional pdf also shifts toward right. Among all these metrics, it seems ccdf is more useful, especially in real time multimedia ....

Jean-Chrysostome Bolot. Characterizing end-to-end packet delay and loss in the Internet. ####### ## #### ##### ########, 2(3):305-323, 1993.

....loss rates. This loss is not catastrophic, as in the days of congestion collapse. Lost packets do, however, waste network bandwidth up to the point where they are discarded. Packet loss rates are hard to characterize globally, but measurements from different times and places are available. Bolot [3] reports loss rates on the order of 9 between France and the US in 1992. Paxson [44] reports a loss rates 2.7 in late 1994 and 5.2 in late 1995, between a variety of pairs of hosts on the Internet. Handley [21] reported loss rates on the order of 10 in 1996 from sites all over the Internet ....

Jean-Chrysostome Bolot. Characterizing end-to-end packet delay and loss in the internet. In Proceedings of ACM SIGCOMM, 1993.

....[Handley, 1997a] analysed loss logs from popular Mbone sessions and concluded that although the deviation of the observed loss patterns from random loss is statistically significant, it is not sufficient to influence the design of a redundancy control algorithm. Similar results are shown in [Bolot, 1993] 3.2. Robust Audio Tool Solutions 45 and [Yajnik et al. 1996] Here, we present a set of measurements that we collected which confirm the above findings and discuss their impact on audio redundancy. Loss statistics were collected using a variety of both unicast and multicast links between UCL, ....

Jean-Chrysostome Bolot. Characterizing end-to-end packet delay and loss in the Internet. Journal of High-Speed Networks, 2(3):305--323, December 1993.

....estimate the document transfer time. We propose to implement two powerful tools for link condition measurement, which we call Bprobe and Cprobe. Bprobe will use a sequence of echo packets to measure the raw or link bandwidth available between client and server (using the principles outlined in [13]) Cprobe will use the information provided by Bprobe, along with a simple stream of probe packets, to discover the amount of intervening traffic between client and server. Together these tools will provide an estimate, not just of round trip time, but also of the maximum available bandwidth and ....

Jean-Chrysostome Bolot. Characterizing end-to-end packet delay and loss in the internet. Journal of High Speed Networks, (2):305--323, 1993.

.... well known effect is illustrated in the familiar diagram shown in Figure 1 Minimum packet spacing at bottleneck link Same spacing is preserved on higher speed links Figure 1: Packet flow through a bottleneck link (adapted from Van Jacobson [10] In addition, Bolot describes the basic effect in [3, 4] and Keshav has used a similar method in networks of Rate Allocating servers [11, 12] The goal of the bprobe tool is to create this condition and to use it to make reliable measurements. In other words, if packets from the probe tool alone are queued at the bottleneck link, then the inter arrival ....

Jean-Chrysostome Bolot. Characterizing End-to-End packet delay and loss in the Internet. Journal of High Speed Networks, 2(3):305--323, 1993.

....and the intercept gives the latency. The line fitting is done by hand and a very large number of packets are needed to make manual fitting possible. An expanded version of that paper gives preliminary suggestions as to how this process can be used to determine the makeup of interfering traffic [1]. However, many obstacles exist to the automatic application of this idea in real networks. Our work overcomes these obstacles and results in a robust procedure that can calculate the bottleneck link speed with minimal overhead. The author in [6] defines the notion of a packet pair, two ....

....spacing will be proportional to the processing time required for the bottleneck router to process the second packet of the pair. This well known effect is illustrated in the familiar diagram shown in Figure 3 (adapted from Van Jacobson [5] In addition, Bolot describes the basic effect in [1, 2] and Keshav has used a similar method in networks of Rate Allocating servers [6] The goal of the bprobe tool is to create this condition and to use it to make reliable measurements. In other words, if packets from the probe tool alone are queued at the bottleneck link, then the inter arrival time ....

Jean-Chrysostome Bolot. Characterizing End-to-End packet delay and loss in the Internet. Journal of High Speed Networks, 2(3):305--323, 1993.

No context found.

Jean-Chrysostome Bolot, \Characterizing End-to-End Packet Delay and Loss in the Internet," Journal of High-Speed Networks, vol. 2, no. 3, pp. 305-323, December 1993.

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