M. Allman and A. Falk, "On the Effective Evaluation of TCP," ACM Comput. Commun. Rev, vol. 29, no. 5, Oct. 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....accuracy of our analytical model is verified through the comparison with measurements using UCB LBNL Network Simulator (ns 2) A. INTRODUCTION A. 1 Related Works Traditional approaches of examining the performance of various aspects of TCP are through simulation, emulation or live measurements [ 14,15]. Only recently several efforts have been directed at analytically characterizing the TCP flow control behavior [3,12,20] Such efforts at modeling are very pertinent to identifying the essential factors that influence or characterize key aspects of TCP performance. In the pioneering work of ....

Mark Allman, Aaron Falk, "On the Effective Evaluation of TCP," ACM Computer Communication Review, 29(5), October 1999.

....access to a shared bottleneck link (N 1 N 2 ) This network model is adequate for addressing the congestion collapse and unfairness issues, although it does not address unfairness caused by differing path lengths. Any reliable TCP simulation study requires an accurate model of the TCP protocol [10]. The simulation model employs a port of BSD 4.4 Lite TCP Reno with 64 kbyte send and receive buffers and 1500 byte MSS. Congestion on the feedback path is not considered in the simulations. PREVENTING CONGESTION COLLAPSE The first scenario illustrates the problem of congestion collapse. In the ....

M. Allman and A. Falk, "On the Effective Evaluation of TCP," ACM Comp. Commun. Rev., vol. 29, no. 5, Oct.

.... advertised by current Internet receivers (less than 64 kbytes) This is supposed to disappear in the future with the trend of increasing the window field in the TCP header and dynamically changing the buffer size allocated to the TCP connection at the receiver [22] It is even recommended in [23] that future studies of TCP congestion control should consider an infinite receiver window. Models that account for the limitation of TCP transmission rate will not be of much importance. Fluid Models vs. Discrete Models Some of the models for TCP assume that the window increases continuously ....

M. Allman and A. Falk, "On the Effective Evaluation of TCP," ACM Comp. Commun. Rev., vol. 29, no. 5, Oct. 1999.

....with the same algorithm. These explanations are further verified by our experiment data in the following. It is known that throughput measurement is a straightforward metric to gauge the performance of TCP. However, it does not capture the detailed behavior of TCP connections. As addressed in [AF00], in real networks, researchers may need to measure loss rate to further evaluate whether modified TCP is aggressive or not. That is, if the improvement of a TCP algorithm is due to its overly aggressive behavior at the expense of other TCP connections, then even if the throughput performance of ....

Mark Allman, Aaron Falk, "On the Effective Evaluation of TCP", SIGCOMM2000"

No context found.

M. Allman and A. Falk, "On the Effective Evaluation of TCP," ACM Comput. Commun. Rev, vol. 29, no. 5, Oct. 1999.

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