S. B. Moon, J. Kurose, and D. Towsley. Correlation of packet delay and loss in the internet. Technical report, Department of Computer Science, University of Massachusetts, Amherst, MA 01003, 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....and loss behavior over the Internet [BOLO93] However, he did not focus on the correlation between an increase in RTT with packet loss. Moon et al. looked at the correlation that exists between a loss conditioned delay (i.e. a packet delay immediately preceding a loss event) and packet loss [MOON99]. Their motivations were similar to ours in that they wanted to see if an endpoint could predict loss. They found a higher level of correlation than we did. Their method utilized a more frequent, one way delay based UDP based probe technique. A further difference is that they do not take into ....

....Conditioned Delay Correlation Metric (LCDC) The loss conditioned delay correlation metric (LCDC) provides a visual indication of the magnitude and time scale of correlation that exists between an increase in tcpRTT samples and loss events. Our algorithm is essentially identical to that used in [MOON99]. Moon et al. define a lag that is used in calculating the average delay conditioned on loss. For each packet loss occurrence in a trace, lag 1 is the tcpRTT sample prior to the transmission of a segment that is lost (and lag 2 is the second tcpRTT sample before the transmission of the ....

[Article contains additional citation context not shown here]

: S. Moon, et. Al., "Correlation of Packet Delay and Loss in the Internet", INFOCOM 1999.

....as the average packet delay and the average packet loss probability. The stringent limitation of the queuing theory is its impossibility to analyze the dynamical behavior of the network. Several measurement based studies suggest that the end toend packet delay of the Internet is quite dynamical [3, 4]. Another approach, being different from queueing theory, should therefore be taken to investigate the packet delay dynamics of the Internet. In this paper, we propose a novel approach to modeling the packet delay dynamics of the Internet. The key idea of our approach is treating the network, ....

S. B. Moon, J. Kurose, P. Skelly, and D. Towsley, "Correlation of packet delay and loss in the Internet," Technical Report 98-11, Jan. 1998.

....the ARX model can capture the round trip time dynamics. In Section 6, we discuss several possible applications of our approach, followed by conclusion of this paper. 2 Related Works In the literature, there have been several measurement based studies regarding the end to end packet delay [3, 4, 8, 9] and the end to end path characteristics [5, 10] In [3] the authors have examined the end to end packet delay and loss behavior in the Internet using small UDP probe packets. In [4] the authors have examined the correlation between packet delay and packet loss experienced by a continuous media ....

....In the literature, there have been several measurement based studies regarding the end to end packet delay [3, 4, 8, 9] and the end to end path characteristics [5, 10] In [3] the authors have examined the end to end packet delay and loss behavior in the Internet using small UDP probe packets. In [4], the authors have examined the correlation between packet delay and packet loss experienced by a continuous media traffic source, based on measurements of per packet delays and packet loss. In [8] a large number of TCP measurements have been used to discuss two estimation problems: estimation of ....

S. B. Moon, J. Kurose, P. Skelly, and D. Towsley, "Correlation of packet delay and loss in the Internet," tech. rep., Department of Computer Science, University of Massachusetts, USA, Jan. 1998.

.... time scales, several milliseconds to seconds, periodic as well as more sophisticated probe streams have been employed to measure bottleneck bandwidth [6] 7] 8] 9] 10] and available bandwidth [7] 10] and to investigate the detailed statistical structure of delay and loss [7] 11] 12] [13], 14] However, at such time scales timing problems in common measurement infrastructures can result in measurement errors of the order of the time intervals one is trying to measure, and the inter departure times of the packet probes one is trying to control. Post processing techniques aimed at ....

S. Moon, J. Kurose, P. Skelly, and D. Towsley, \Correlation of packet delay and loss in the internet," Tech. Rep. TR 98-11, Department of Computer Science, University of Massachusetts, January 1998.

....to the same block are received. The loss pattern, whether the original one or the nal losses after FEC, a ects audio video quality and e ectiveness of loss concealment [9] To the end user, burstiness in late losses has no di erence from network losses. Finally, as reported by Moon et al. [13] there is inter dependency between delay and loss. It means late losses and network losses may merge into longer nal loss bursts. This e ect is shown later in this paper, ment loss concealFEC recovery unrecovered losses packets with playout delay late loss added more playout delay ....

....is determined by the nal loss patterns (FLPs) after playout delay control and optionally FEC. FEC can signi cantly change the loss rate and its distribution. The FLP could be even burstier due to the inter dependency between delay and loss, that is, a packet loss is often preceded by high delays [13]. Finally, FEC coupled with playout delay adjustment complicates the FLP even further because a recovery time longer than the playout delay is equivalent to a late loss. We run a set of playout control simulations based on our Internet packet traces. The FLP is indeed bursty and also ts with the ....

Sue B. Moon, Jim Kurose, Paul Skelly, and Don Towsley. Correlation of packet delay and loss in the internet. Technical report, University of Massachusetts, January 1998.

....bestimmte Spracheigenschaften aus, um Audiopakete ber ein verlustbehaftetes, paketvermittelndes und auf dem Best Effort Prinzip basierendes Netzwerk zu bertragen. Obwohl Paketverzgerung und Paketverlust einen schlechten Einflu auf die Qualitt der Sprachbertragung haben und auerdem korrelieren [SKBD98], beschrnkt sich unsere Arbeit nur auf die Probleme der Paketverluste 1 . In dieser Arbeit leisten wir zwei Beitrge zu dem Forschungsgebiet der Sprachbertragung ber das Internet: Verbesserung des Verfahrens Adaptive Packetization and Concealment (AP C) Sann98a] Sann98b] AP C paketisiert ....

....Nguyen Tuong Long Le Abstracts This thesis exploits certain properties of speech signals to transmit audio packets over a best effort packet switched network. Although both packet delay and packet loss have an adverse impact on the quality of speech transmission and furthermore are correlated [SKBD98], this thesis only concentrates on the problems of packet loss 1 . In this thesis, we present two contributions to the research topic of speech transmissions over the Internet: Improvements of the Adaptive Packetization and Concealment (AP C) Sann98a] Sann98b] AP C packetizes and ....

[Article contains additional citation context not shown here]

S. B. Moon, J. Kurose, P. Skelly, D. Towsley. Correlation of Packet Delay and Loss in the Internet. Technical Report, Department of Computer Science, University of Massachusetts at Amherst.

....the orders of the ARX model are also discussed. In Section 6, we discuss several possible applications of our approach. Section 7 concludes this paper with a few remarks. 3 2. Related Works In the literature, there have been several measurement based studies regarding the endto end packet delay [4, 5, 9, 10] and the end to end path characteristics [6, 11] In [4] the authors have examined the end to end packet delay and loss behavior in the Internet using small UDP probe packets. In [5] the authors have examined the correlation between packet delay and packet loss experienced by a continuous media ....

....the literature, there have been several measurement based studies regarding the endto end packet delay [4, 5, 9, 10] and the end to end path characteristics [6, 11] In [4] the authors have examined the end to end packet delay and loss behavior in the Internet using small UDP probe packets. In [5], the authors have examined the correlation between packet delay and packet loss experienced by a continuous media tra#c source based on measurements of per packet delays and packet loss. In [9] a large number of TCP measurements have been used to discuss two estimation problems: estimation of ....

[Article contains additional citation context not shown here]

S. B. Moon, J. Kurose, P. Skelly, and D. Towsley, "Correlation of packet delay and loss in the Internet," tech. rep., Department of Computer Science, University of Massachusetts, USA, Jan. 1998.

....voice streams have low data rates (in the order of tens of Kbit s) and exhibit low burstiness. Because of these stringent requirements and particular characteristics, voice traffic should be treated differently than other traffic in the network. In fact, measurements on the Internet [28] as well as simulation studies [22] have shown that mixing voice traffic with both traditional TCP data traffic and UDP VBR video traffic can lead to either low average link utilization if delay requirements are met, or larger than desired delay for voice. Accordingly, allowing a mixture of voice ....

S. Moon, J. Kurose, P. Skelly, and D. Towsley, "Correlation of Packet Delay and Loss in the Internet," Technical Report 98-11, Department of Computer Science, University of Massachussetts, Amherst, Jan. 1998.

....model can capture the round trip time dynamics. In Section 7, we discuss several possible applications of our approach. Section 8 concludes this paper with a few remarks. 2. Related Works In the literature, there have been several measurementbased studies regarding the end to end packet delay [4, 5, 9, 10] and the end to end path characteristics [6, 11] In [4] the authors have examined the end to end packet delay and loss behavior in the Internet using small UDP probe packets. In [5] the authors have examined the correlation between packet delay and packet loss experienced by a continuous media ....

....In the literature, there have been several measurementbased studies regarding the end to end packet delay [4, 5, 9, 10] and the end to end path characteristics [6, 11] In [4] the authors have examined the end to end packet delay and loss behavior in the Internet using small UDP probe packets. In [5], the authors have examined the correlation between packet delay and packet loss experienced by a continuous media tra#c source, based on measurements of per packet delays and packet loss. In [9] a large number of TCP measurements have been used to discuss two estimation problems: estimation of ....

S. B. Moon, J. Kurose, P. Skelly, and D. Towsley, "Correlation of packet delay and loss in the Internet," tech. rep., Department of Computer Science, University of Massachusetts, USA, Jan. 1998.

....observed loss. The temporal correlation of packet loss has been noted by a number of authors. Bolot [5] noted that packet loss is not independent (it is more likely that a packet is dropped if the previous packet was also lost) and derived a simple Bernoulli model for such loss. More recent work [31, 20] notes that this model is not sufficient in many cases, and that higher order Markov models are more accurate. Correlation is also noticable at longer time scales. For example, Handley [10] and Bolot [5] have noted bursts of loss with a 30 second period (possibly due to router bugs) and the ....

S. Moon, J. Kurose, P. Skelly and D. Towsley. Correlation of Packet Delay and Loss in the Internet. Technical Report 98-11, Department of Computer Science, University of Massachusetts, Amherst, MA 01003, USA.

....cycles of delay and loss matches well with each other [21] This result is intuitive because losses generally occur because of congestion, and congestion leads to higher delays. On a short time scale of around 100ms, however, the inter dependencies of delay and loss are quite di erent. Moon et al. [27] reports that loss is not very sensitive to (either long or short) delay, either backward or forward in time, and delay is quite sensitive to loss, generally a loss appears after several high delays, and it is followed by low delay. The second result is intuitive because a loss in the Internet is ....

....(a high delay will follow by a low delay) For di erent lags, the value will di er. In general, it is bigger when l = 0, and drops to 0 as l increases. This matches our intuition, that adjacent delays are correlated, but delays of packets far apart are independent. However, it is shown in [27] that sometimes auto correlation can go up periodically. They conjectured it is caused by TCP or some other synchronization e ect in the network. The auto correlation value is a good indicator of dependency, but it is dicult to quantify the dependency using this metric. Therefore, we introduce a ....

Sue B. Moon, Jim Kurose, Paul Skelly, and Don Towsley. Correlation of packet delay and loss in the internet. Technical report, University of Massachusetts, January 1998.

....congestion sampling algorithm in predicting future packet loss events. Several previous studies are relevant. Bolot [6] studied end to end packet delay and loss behavior over the Internet. However, he did not focus on the correlation between an increase in RTT with packet loss. Moon et al. [13] looked at the correlation that exists between a loss conditioned delay (i.e. a packet delay immediately preceding a loss event) and packet loss. Their motivations were similar to ours in that they wanted to see if an endpoint could predict loss. They found a higher level of correlation than we ....

....Conditioned Delay Correlation Metric (LCDC) The loss conditioned delay correlation metric (LCDC) provides a visual indication of the magnitude and time scale of correlation that exists between increases in tcpRTT samples and loss events. Our algorithm is essentially identical to that used in [13]. Moon et al. define a lag that is used in calculating the average delay conditioned on loss. For each packet loss occurrence in a trace, lag 1 is the tcpRTT sample prior to the transmission of a segment that is lost (and lag 2 is the second tcpRTT sample before the transmission of the lost ....

[Article contains additional citation context not shown here]

S. Moon, et. Al., "Correlation of Packet Delay and Loss in the Internet", INFOCOM 1999.

....cycles of delay and loss matches well with each other [21] This result is intuitive because losses generally occur because of congestion, and congestion leads to higher delays. On a short time scale of around 100ms, however, the inter dependencies of delay and loss are quite di erent. Moon et al. [27] reports that loss is not very sensitive to (either long or short) delay, either backward or forward in time, and delay is quite sensitive to loss, generally a loss appears after several high delays, and it is followed bylow delay. The second result is intuitive because a loss in the Internet is ....

....correlated (a high delay will followbyalow delay) For di erent lags, the value will di er. In general, it is bigger when # = 0, and drops to 0 as # increases. This matches our intuition, that adjacent delays are correlated, but delays of packets far apart are independent. However, it is shown in [27] that sometimes auto correlation can go up periodically. They conjectured it is caused by TCP or some other synchronization e ect in the network. The auto correlation value is a good indicator of dependency, but it is dicult to quantify the dependency using this metric. Therefore, weintroduce a ....

Sue B. Moon, Jim Kurose, Paul Skelly, and Don Towsley. Correlation of packet delay and loss in the internet. Technical report, University of Massachusetts, January 1998.

.... degree of positive correlation (i.e. a loss or excessive delay observed by a packet increases the likelihood that a later packet will be lost or experience a large delay) However, in general, the degree of correlation decreases as the time between the packets transmissions is increased [17] [18]. The losses or delays experienced by two packets that do not share the same POC will exhibit little or no correlation. Our idea is to measure the correlation between pairs of packets both within a ow, and between ows. We choose the pairs between ows such that if the POC for the ows is ....

S. Moon, J. Kurose, and D. Towsley. Correlation of Packet Delay and Loss in the Internet. Technical report, University of Massachusetts, CMPSCI 98-11, January 1998.

.... POC exhibit some degree of correlation (i.e. a loss or excessive delay observed by a packet increases the likelihood that a later packet will be lost or experience a large delay) However, in general, the degree of correlation decreases as the time between the packets transmissions is increased [11, 12]. The losses or delays experienced by two packets that do not share the same POC will exhibit little or no correlation. Our idea is to measure the correlation between pairs of packets both within a flow, and between flows. We choose the pairs between flows such that if the POC for the flows is ....

S. Moon, J. Kurose, and D. Towsley. Correlation of Packet Delay and Loss in the Internet. Technical report, UMass CMPSCI Technical Report 98-11, January 1998.

No context found.

S. Moon, J. Kurose, and D. Towsley. Correlation of Packet Delay and Loss in the Internet. Technical report, UMass CMPSCI Technical Report 98-11, January 1998.

No context found.

S. Moon, J. Kurose, P. Skelly and D. Towsley, "Correlation of Packet Delay and Loss in the Internet" Tech. Report University of Massachussets at Amherst,1999.

No context found.

S. B. Moon, J. Kurose, and D. Towsley. Correlation of packet delay and loss in the internet. Technical report, Department of Computer Science, University of Massachusetts, Amherst, MA 01003, 1998.

No context found.

S.B. Moon, J. Kurose, P. Skelly, and D. Towsley, "Correlation of packet delay and loss in the Internet, " Technical Report UM-CS-1998-011, University of Massachusetts, March 1998.

No context found.

S. B. Moon, J. Kurose, P. Skelly, and D. Towsley, "Correlation of packet delay and loss in the Internet," Tech - nical Report 98-11, Jan. 1998.

No context found.

S. B. Moon, J. Kurose, P. Skelly, and D. Towsley, "Correlation of packet delay and loss in the Internet, " tech. rep., Department of Computer Science, University of Massachusetts, USA, Jan. 1998.

No context found.

S. Moon, J. Kurose, P. Skelly and D. Towsley. Correlation of Packet Delay and Loss in the Internet. Technical Report 98-11, Department of Computer Science, University of Massachusetts, Amherst, MA 01003, USA.

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