Jiang, H. Schulzrinne. "Modeling of packet loss and delay and their effect on real-time multimedia service quality", in Proceedings of NOSSDAV '2000.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....10 byte packet standard, and assume that the position of the missing packets can be identified by the sequence ID number which is included in the header of each packet. A. Channel Model Most real communication channels exhibit burstiness. Such channels can be modeled by a 2 state Markov model [41], also known as a Gilbert model. In Fig. 3, is the probability that the next packet is lost, provided the previous one has arrived; is the probability that the next packet is not lost, given that the previous one was lost. If , the Gilbert model reduces to a Bernoulli model. The parameter can be ....

....one was lost. If , the Gilbert model reduces to a Bernoulli model. The parameter can be seen as controlling the burstiness of packet losses. Although our FEC assignment was optimized with an exponential packet loss probability mass function, we ran tests under the loss conditions reported in [42] [41]. These conditions are summarized in Table IV, where is the mean loss probability, i.e. the parameter of the exponential model used in the FEC assignment exponential model, and is the conditional loss probability, conditioned on the event that the previous packet was lost. Fig. 3. Gilbert ....

W. Jiang and H. Schulzrinne, "Modeling of packet loss and delay and their effect on real-time multimedia service quality," in Proc. 10th Int. Workshop Network and Operations System Support for Digital Audio and Video, June 2000.

....varies from 10 to 0. We assume ##### # # # ############ for each link, which corresponds to 5 end to end average packet loss rate for 10 links, and average burst length of 1. 25 packets which is the longest average burst length (for 30 msec sampling) that we are aware of in the literature [9, 10]. In these plots, the SD performance is constant because SD is sent along a single path that is independent of jointness. For maximally disjoint paths, the MD distortion is minimized and it significantly outperforms SD. As the jointness of the paths increases, the MD distortion increases ....

....burst length corresponds to an increased probability of simultaneous loss of both descriptions for MD, corresponding to significantly higher distortion. It is worth stressing that the longest expected end to end burst length that has been experimentally measured and cited in the literature [9, 10] that we are aware of is 1.25 packets, and in the region of expected burst length less than 1.25 packets, the expected distortion of MD is less than that of SD. SD distortion depends primarily on the loss rate and to a lesser extent on the expected burst length (relatively small spread in Fig 3 ....

J. Wenyu and H. Schulzrinne, "Modeling of packet loss and delay and their effects on real-time multimedia service quality," NOSSDAV'00.

....calls from transport level measurable quantities. In Section 5 we finally discuss our findings while Section 6 presents some concluding remarks. 2 Related work Past literature on end to end Internet measurements has often focused on the study of network loss patterns and delay characteristics [5, 6, 14, 24, 22]. For example, Kostas [16] studied the feasibility of real time voice over the Internet and discussed measured delay and loss characteristics. In order to evaluate the quality of Internet Telephony, 12] provided network performance data (in terms of delay and losses) collected from a wide range ....

W. Jiang and H. Schulzrinne. Modeling of packet loss and delay and their effect on real-time multimedia service quality. In Proceedings of ACM NOSSDAV, June 2000.

....model requires memory of the state of the previous packet (lost or not lost) Since this model only retains memory of the state of the last packet, it does not contain the granularity necessary to accurately model the multiple packet losses that occur within a packet loss burst. Moreover, Jiang [JiSc2000] found that this model underestimated the probability of consecutive packet losses, and thus could not accurately model packet bursts. For these reasons, this model was rejected. To retain a limited amount of memory of the gaps between packet loss events, a 6 state Markov chain was chosen for the ....

W. Jiang, H. Schulzrinne, "Modeling of Packet Loss and Delay and Their Effect on Real-Time Multimedia Service Quality", 10th Intemational Workshop on Network and Operating System Support for Digital Audio and Video. June, 2000, Chapel Hill, NC.

....calls from transport level measurable quantities. In Section 5 we finally discuss our findings, while Section 6 presents some concluding remarks. 2. RELATED WORK Past literature on end to end Internet measurements has often focused on the study of network loss patterns and delay characteristics [6, 8, 16, 26, 24]. For example, Kostas [18] studied the feasibility of real time voice over the Internet and discussed measured delay and loss characteristics. In order to evaluate the quality of Internet Telephony, 14] provided network performance data (in terms of delay and losses) collected from a wide range ....

W. Jiang and H. Schulzrinne. Modeling of packet loss and delay and their effect on real-time multimedia service quality. In Proceedings of ACM NOSSDAV, June 2000.

....The WER for this test set is 22.9 , comparable to what other labs are reporting [9] Although the FEC assignment was optimized with an exponential packet loss probability distribution function, most real channels exhibit burstiness. We have modeled such channels by using a 2 state Markov model [6], known also as a Gilbert model. In Figure 4, p is the probability that the next packet is lost, provided the previous one has arrived; q is the opposite. If p q = 1 the Gilbert model reduces to a Bernoulli model. q can be seen as controlling the burstiness of packet losses. The tests were run ....

....Figure 4, p is the probability that the next packet is lost, provided the previous one has arrived; q is the opposite. If p q = 1 the Gilbert model reduces to a Bernoulli model. q can be seen as controlling the burstiness of packet losses. The tests were run under the loss conditions reported in [12, 6] and are summarized in Table 2, where ulp = p= p q) is the unconditional loss probability which was used in the FEC assignment exponential model, and clp = 1 q is the loss probability, conditioned on the event that the previous packet was lost. 0 (non loss) 1 (loss) 1 p 1 q p q Figure 4: ....

W. Jian and H. Schulzrinne. Modeling of packet loss and delay and their effect on real-time multimedia service quality. In The 10th International Workshop on Network and Operations System Support for Digital Audio and Video, June 2000.

....model must capture this runlength dependence. For a single path, the end to end loss process is commonly represented by the Gilbert model with transition probabilities ### ##, where # is the probability of going from no loss (0) to loss (1) and # is the probability of going from loss to no loss [7, 8]. The Gilbert model only models the loss process of a path but not the distortion when video is transmitted over that path. One distortion model for transmitting SD video over a single path is shown in Fig. 2, where the states denote the number of consecutive losses in the immediate past. The ....

....disjoint links therefore varies from 8 to 0. We assume ##### ## # ########### for each link, which corresponds to 5 end to end average packet loss for 8 links, and average burst length of 1. 25 packets which corresponds to the longest (for 30 msec sampling) that we are aware of in the literature [7, 8]. Different fixed path diversities, vary end to end loss rate. Figures 5 shows MD vs SD performance for three different degrees of path diversity as we vary the average packet loss rate per link from 0 to 5 , and ## # ##. The three topologies examined are: 1) completely disjoint, topology of ....

J. Wenyu and H. Schulzrinne, "Modeling of packet loss and delay and their effects on real-time multimedia service quality," NOSSDAV, 2000.

....to no loss (0) The Gilbert model is widely used to model bursty traffic for its simplicity and mathematical tractability. While prior work modeled end to end packet loss across a single path APOSTOLOPOULOS, WONG, TAN, AND WEE: ON MULTIPLE DESCRIPTION STREAMING WITH CDNS 5 in the Internet [14] [15], we propose single link models which are then used to develop end to end loss models. Second, we also assume that each link can be modeled as independent. A path is modeled as the concatenation of a number of bursty single links. When each bursty single link in a path of # links is modeled as a ....

....varies from 8 to 0. For this plot we assumed ## # ## # # # ########## for each link, where the # # corresponds to 5 end to end average packet loss for 8 links, and # # # ## corresponds to the longest average burst length (assuming 30 msec sampling) that we are aware of in the literature [14] [15]. 0 1 p q 1 q 1 p 2 3 1 q 1 q qq D drop1 D drop2 D drop3 D drop3 D no loss D rec2 D rec1 D rec3 Fig. 3. Expected SD video quality is estimated from this model, where the four states identify the burst length and where the transition probabilities are labeled as well as the ....

[Article contains additional citation context not shown here]

J. Wenyu and H. Schulzrinne, "Modeling of packet loss and delay and their effects on real-time multimedia service quality," NOSSDAV, 2000.

....the question of independence at larger time scales. We introduce a simple refinement to such characterizations that allows us to identify these correlations as due to backto back loss rather than nearby loss, and we relate the result to the extended Gilbert loss model family [Gi60] SCK00] [JS00]. We do so by considering not the loss process itself, but the loss episode process, i.e. the time series indicating when a series of consecutive packets (possibly only of length one) were lost. For loss processes, we expect congestion induced events to be clustered in time, so to assess ....

W. Jiang and H. Schulzrinne, "Modeling of Packet Loss and Delay and Their Effect on Real-Time Multimedia Service Quality," Proc. NOSSDAV

....the question of independence at larger time scales. We introduce a simple refinement to such characterizations that allows us to identify these correlations as due to backto back loss rather than nearby loss, and we relate the result to the extended Gilbert loss model family [Gi60] SCK00] [JS00]. We do so by considering not the loss process itself, but the loss episode process, i.e. the time series indicating when a series of consecutive packets (possibly only of length one) were lost. For loss processes, we expect congestion induced events to be clustered in time, so to assess ....

W. Jiang and H. Schulzrinne, "Modeling of Packet Loss and Delay and Their Effect on Real-Time Multimedia Service Quality," Proc. NOSSDAV

No context found.

Jiang, H. Schulzrinne. "Modeling of packet loss and delay and their effect on real-time multimedia service quality", in Proceedings of NOSSDAV '2000.

No context found.

W. Jiang and H. Schulzrinne, "Modeling of packet loss and delay and their effect on real-time multimedia service quality," NOSSDAV 2000.

No context found.

W. Jiang and H. Schulzrinne, "Modeling of packet loss and delay and their effect on real-time multimedia service quality," in Proc. of NOSSDAV, 2000.

No context found.

W. Jiang and H. Schulzrinne, "Modeling of packet loss and delay and their effect on real-time multimedia service quality," in Proc. Int. Workshop Network and Operating Systems Support for Digital Audio and Video NOSSDAV, Chapel Hill, NC, June 2000.

No context found.

W. Jiang and H. Schulzrinne, "Modeling of Packet Loss and Delay and Their Effect on Real-Time Multimedia Service Quality," 10th International Workshop on Network and Operating System Support for Digital Audio and Video, 2000.

No context found.

W. Jiang and H. Schulzrinne. Modeling of packet loss and delay and their effect on real-time multimedia service quality. In Proceedings of NOSSDAV, 2000.

No context found.

W. Jiang and H. Schulzrinne, "Modeling of packet loss and delay and their effect on real-time multimedia service quality," in Proc. NOSSDAV, Chapel Hill, NC, June 2000.

No context found.

W. Jiang and H. Schulzrinne, "Modeling of packet loss and delay and their effect on real-time multimedia service quality," in Proc. Int. Workshop Network and Operating Systems Support for Digital Audio and Video NOSSDAV, Chapel Hill, NC, June 2000.

No context found.

W. Jiang et al., "Modeling of packet loss and delay and their effect on real-time multimedia service quality", Proceedings of NOSSDAV 2000.

No context found.

W. Jiang and H. Schulzrinne, "Modeling of packet loss and delay and their effect on real-time multimedia service quality," in Proc. of NOSSDAV, 2000.

No context found.

W. Jiang and H. Schulzrinne, "Modeling of Packet Loss and Delay and Their Effects on Real-time Multimedia Service Quality," NOSSDAV, 2000.

No context found.

W. Jiang and H. Schulzrinne. Modeling of packet loss and delay and their effect on real-time multimedia service quality. In Proceedings of ACM NOSSDAV, June 2000.

No context found.

W. Jiang and H. Schulzrinne. Modeling of packet loss and delay and their effect on real-time multimedia service quality. In Proceedings of NOSSDAV, 2000.

No context found.

W. Jiang and H. Schulzrinne. Modeling of Packet Loss and Delay and Their Effect on Real-Time Multimedia Service Quality. In The 10th International Workshop on Network and Operating System Support for Digital Audio and Video (NOSSDAV 2000), June 2000.

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