A. Fei, G. Pei, R. Liu, and L. Zhang, "Measurements on Delay and Hop-Count of the Internet," Proc. IEEE GLOBECOM '98---Internet Mini-Conf., 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....two (arbitrary) routers in Internet. Therefore, understanding the end to end delay components in the (Inter)network is crucial to assess the possibility and the level of quality of service (QoS) of real time services and to improve the current Internet architecture. Several papers [4] 16] [10] and [21] report end to end delay measurements, mostly based on round trip time (RTT) Difficulties with clock synchronization, asymme tries in the one way and return path and path variations during the measurement limit RTT based measurements. To surmount these problems and because large ....

A. Fei, G. Pei, R. Liu and L. Zhang, "Measurements on delay and hop-count of the Internet", Proc. of the IEEE Global Internet '98, December 1998.

.... we present results primarily for the 700 node scenario (1 source, 100 repositories and 600 routers) In such a network, an update from one repository (or source) to another traverses around 10 hops on average, compared to the 18 hops reported based on measurements done on the internet in [9]. Results for other network sizes are briefly discussed in Section 6.3.5. Our experiments use node node communication delays derived from a heavy tailed Pareto [19] distribution: w . is given by . being the mean . is the minimum delay a link can have. For our ....

.... . w . is given by . being the mean . is the minimum delay a link can have. For our experiments, was 15 ms (milli secs) and . was 2 ms. As a result, the average nominal node node delay in our networks was around 20 30 ms. This is lower than the delays reported in [9]. We also experimented with higher network delays in Section 6.3.2 and show that the gain in the fidelity using cooperative dissemination is even more significant for higher delays. Unless otherwise specified, computational delay incurred at a repository to disseminate an update to a dependent is ....

A. Fei, G. Pei, R. Liu, and L. Zhang. Measurements on delay and hop-count of the internet. In IEEE GLOBECOM'98 - Internet Mini-Conference, 1998.

....the logarithmic transmission success probability function (slog) instead of the loss probability to avoid complex composition rules. slog(p) jlog(1 L(p) j: 1) Furthermore, we assume that routes can not have more than 90 of loss probability, that is, 0 slog(p) 1. As ms unit is appropriate [21] to represent delay, we assume that delay is an integer. As a consequence, we have a simple single metric representation of delay and loss, where the integer part is delay and the decimal part represents loss. This property allows the construction of simple algorithms capable of nding paths ....

A. Fei, G. Pei, R. Liu, and L. Zhang, \Measurements on delay and hop-count of the Internet," in IEEE GLOBECOM'98 - Internet Mini-Conference, 1998.

....loss probability, the absolute value of the logarithmic transmission success probability function (slog(p) jlog(1 L(p) j) is used to avoid complex composition rules. Furthermore, we assume that routes can not have more than 90 of loss probability, so 0 jslogj 1. As ms unit is appropriate [18] to represent delay, we assume that delay is an integer. As a consequence, we have a simple single metric representation of delay and loss, where the integer part is delay and the decimal part represents loss. This property allows the construction of a simple heuristic. Let G = N; A) be a network ....

A. Fei, G. Pei, R. Liu, and L. Zhang, \Measurements on delay and hop-count of the Internet," in IEEE GLOBECOM'98 - Internet Mini-Conference, 1998.

....logarithmic transmission success probability function (slog) instead of the loss probability to avoid complex composition rules. slog(p) jlog(1 L(p) j: 3.1) Furthermore, we assume that routes can not have more than 90 of loss probability, that is, 0 slog(p) 1. As ms unit is appropriate [58] to represent delay, we assume that delay is an integer. As a consequence, we have a simple single metric representation of delay and loss, where the integer part is delay and the decimal part represents loss. This property allows the construction of simple algorithms capable of nding paths ....

A. Fei, G. Pei, R. Liu, and L. Zhang, Measurements on delay and hop-count of the Internet, in IEEE GLOBECOM'98 - Internet Mini-Conference, 1998.

....Florham Park, NJ. Email: walter research.att.com. methodologies can often elicit certain responses to that stimulus from individual network components or from the network. In turn, these responses are used to infer performance metrics such as bulk throughput [29] packet delay [10] [18], 32] 35] packet loss [32] 35] 37] bottleneck link bandwidth [10] 14] 25] 32] 35] link bandwidths along the end to end path [17] 24] 26] hop counts [18] or other path characteristics [31] 32] 35] 23] 9] 30] Active measurements cover only those parts of the ....

....network. In turn, these responses are used to infer performance metrics such as bulk throughput [29] packet delay [10] 18] 32] 35] packet loss [32] 35] 37] bottleneck link bandwidth [10] 14] 25] 32] 35] link bandwidths along the end to end path [17] 24] 26] hop counts [18], or other path characteristics [31] 32] 35] 23] 9] 30] Active measurements cover only those parts of the network that are being probed, do not generally deal with application performance, and ISPs are typically restricted to use them within their network. In addition, active ....

A. Fei, G. Pei, R. Liu, and L. Zhang. Measurements on delay and hop-count of the Internet. Proc. IEEE Global Internet'98.

....Instead of loss probability, the algorithm uses the logarithm of transmission success probability function (slog) to avoid complex composition rules. Furthermore, we assume that routes can not have more than 90 of loss probability, that is, 0 jslogj 1. As ms unit is likely to be appropriate [22] to represent delay, we assume that delay is an integer. As a consequence, we have a simple single metric representation of delay and loss, where the integer part is delay and the decimal part represents loss. The usual approach in current hop by hop routing algorithms is to compute the best path ....

A. Fei, G. Pei, R. Liu, and L. Zhang, \Measurements on delay and hop-count of the Internet," in IEEE GLOBECOM'98 - Internet Mini-Conference, 1998.

....Instead of loss probability, the algorithm uses the logarithm of transmission success probability function (slog) to avoid complex composition rules. Furthermore, we assume that routes can not have more than 90 of loss probability, that is, 0 jslogj 1. As ms unit is likely to be appropriate [20] to represent delay, we assume that delay is an integer. As a consequence, we have a simple single metric representation of delay and loss, where the integer part is delay and the decimal part represents loss. This property allows us to construct a simple heuristic capable of nding feasible ....

A. Fei, G. Pei, R. Liu, and L. Zhang, \Measurements on delay and hop-count of the Internet," in IEEE GLOBECOM'98 - Internet Mini-Conference, 1998.

No context found.

A. Fei, G. Pei, R. Liu, and L. Zhang, "Measurements on Delay and Hop-Count of the Internet," Proc. IEEE GLOBECOM '98---Internet Mini-Conf., 1998.

No context found.

A. Fei, G. Pei, and L. Zhang. Measurements on delay and hop-count of the internet. IEEE GLOBECOM'98 - Internet Mini-Conference, 1998.

No context found.

A. Fei, G. Pei, R. Liu, and L. Zhang. Measurements on delay and hop-count of the internet. In IEEE GLOBECOM'98 - Internet Mini-Conference, 1998.

No context found.

A. Fei, G. Pei, R. Liu, and L. Zhang, Measurements on Delay and Hop-count of the Internet, Proc. of the IEEE Global Internet, December 1998.

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