M. Handley and S. Floyd, "TCP-friendly simulations," unpublished.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....by receiver s window size, and explicit modeling of TCP s fast retransmit mechanisms. While previously proposed models are valid only for loss rates below 5 [34] our model is valid over a almost the entire range of packet loss rates. Our own measurements, as well as those of independent studies [3, 5, 16, 52] show that our model provides a good t to observed data under a wide variety of network conditions. Validation of the model by Internet measurements: To validate our model, we gathered data by tracing the packet level dynamics of over several TCP connections over the Internet. In all, we ....

....values than the TD Only model. Our approximate model also generally provides more accurate predictions than the TD Only model, and is quite close to the predictions made by the full model. Empirical and simulation studies of the model proposed in this chapter have also been presented in [3, 5, 16, 52]. These studies have found that the model provides a good t to the observed send rate of TCP connections under a wide variety of network conditions. 39 2.4 Discussion of the model and the experimental results In this section, we discuss various simplifying assumptions made while constructing ....

Handley, M., and Flyod, S. TCP-Friendly Simulations. Unpublished Manuscript.

....W a ; Wmax ; p 1 and p 2 . A recent work [22] models the behavior of TCP Reno as a Markov process for a single service, where all the packets incur the same loss rate. This model has been shown to be reasonably accurate when compared to the measurements over the Internet [21] and with simulation [15]. We modify the above Markov model to account for the two regimes of packet loss that we have assumed for in profile and out profile packets. Each state Q i in the Markov chain represents the current window size W i , the number of packets that are lost in the previous round L i , and whether or ....

M. Handley, S. Floyd. TCP-Friendly Simulations, Unpublished Work.

....W a ; W max ; p 1 and p 2 . A recent work [22] models the behavior of TCP Reno as a Markov process for a single service, where all the packets incur the same loss rate. This model has been shown to be reasonably accurate when compared to the measurements over the Internet [21] and with simulation [15]. We modify the above Markov model to account for the two regimes of packet loss that we have assumed for in profile and out profile packets. Each state Q i in the Markov chain represents the current window size W i , the number of packets that are lost in the previous round L i , and whether or ....

M. Handley, S. Floyd. TCP-Friendly Simulations, Unpublished Work.

.... baseline protocols that try to mimic the major features of TCP congestion control algorithm without providing reliable delivery have been proposed [7, 17, 20, 24] Some ongoing work, based partially on our ideas, with a focus on formula based multicast congestion control, is also reported in [5, 6]. We discuss some of these protocols and their limitations in the next section. We believe there are several advantages to taking a formula based approach towards developing a TCP friendly congestion control scheme. First, a formula based approach is flexible. By changing the formula, one can ....

....more complex answers than the three simple strategies described here. The recomputation interval must be short enough to allow TFRCP to be responsive, while at the same time it must be large enough to allow the loss rate measurements to be meaningful. This question is currently under research [5, 6]. Thus, it is appropriate to restrict the baseline protocol described here to the simplest strategy. 0 0.5 1 1.5 2 2.5 3 10 15 20 25 30 35 40 45 50 Friendliness (F) Total Flows Constant Bottleneck Bandwidth: S1 M=2 sec M=3 sec M=4 sec M=5 sec (a) Strategy S1 0 1 2 3 4 5 6 7 10 15 20 25 30 35 ....

M. Handley and S. Flyod. TCP-Friendly Simulations. Unpublished Manuscript.

No context found.

M. Handley and S. Floyd, "TCP-friendly simulations," unpublished.

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