S. Biaz and N. Vaidya, "Using End-to-end Statistics to Distinguish Congestion and Corruption Losses: A Negative Result," Texas A&M University, Technical Report 97-009, August 18, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....environment, a natural issue is the achievable improvement when the sender is able to distinguish between congestion and corruption losses. In other words, for each packet loss, the TCP sender knows if the loss occurred due to congestion or due to corruption loss over the wireless link. A study [BV97] shows that the improvement depends on the ratio of congestion and corruption loss probabilities. The result is obtained from experiments and theoretical approximations as follows. A simple approximation for long range throughput # of TCP Reno is from [MSMO97] where ### is the Maximum Segment ....

....TCP Reno is introduced that has perfect knowledge of the reason for a packet loss, and thus halves its congestion window only for congestion losses. The approximation of long range throughput for Ideal Reno is # # # The authors do not give a valid range of parameters for the estimation [BV97] We believe that the approximation is only valid when # # and # # are of a few per cent. The improvement of Ideal TCP Reno over TCP is approximated as # ##### ##### # #### ## # . We see that # # ## # is the main factor of how much better Ideal TCPReno can perform. The secondary factor ....

[Article contains additional citation context not shown here]

S. Biaz and N. Vaidya. Using end-to-end statistics to distinguish congestion and corruption losses: A negative result. Technical Report TR97-009, Texas A&M University, August 9, 1997.

....without adversely affecting other concurrent TCP connections in the same network. Such improvement is obtained not only in wireless networks but also in asymmetric networks and in high latency networks. By contrast, many algorithms proposed by others have never been tested over real networks [BV97, BV98, BV98a]. 3) Implementation of Veno TCP in NetBSD1.0. Veno has been implemented in the kernel of NetBSD1.0. Similar to TCP Vegas, Veno maintains a timestamp queue for all outstanding segments to perform more accurate RTT calculation and obtain more accurate estimation of network states. Overall, only ....

Biaz, S., Vaidya, N., "Using End-to-end Statistics to Distinguish Congestion and Corruption Lossses: A Negative Result," Texas A&M University, Technical Report 97009, August 18, 1997."

....suggested statistical methods for congestion avoidance [Jain89, WC91, VEGAS] A natural extension of these heuristics would enable a sender to distinguish between losses caused by congestion and other causes. Unfortunately, it does not seem like these sender based heuristics are at all reliable [BV97, BV98]. Fortunately, under selected conditions recent results using packet inter arrival times measured at the receiver are encouraging [BV98a] 4.7.2 With Explicit Notification Explicit notification from the network can make it very easy to determine when a loss is not due to congestion, so as to ....

Biaz, S., Vaidya, N., "Using End-to-end Statistics to Distinguish Congestion and Corruption Lossses: A Negative Result," Texas A&M University, Technical Report 97-009, August 18, 1997.

No context found.

S. Biaz and N. Vaidya, "Using End-to-end Statistics to Distinguish Congestion and Corruption Losses: A Negative Result," Texas A&M University, Technical Report 97-009, August 18, 1997.

....Ideally, it would help if a sender could di erentiate between packet losses due to congestion from the packet losses due to wireless transmission errors using some end to end technique. Attempts to apply heuristics to distinguish between congestion and transmission errors have not been successful [1][2] 3] Other techniques of using Performance Enhancing Proxies [4] at the boundary of wireless networks require TCPlevel awareness by the intermediate nodes. A Random Early Detection(RED) 5] enabled router detects congestion before the bu er over ows, based on a running average queue size, and ....

S. Biaz and N. Vaidya, \Using end-to-end statistics to distinguish congestion and corruption losses: a negative result," Technical Report 97-009, Texas A&M University, August 1997.

....This suggests that the three predictors cannot correctly detect queue build up, and hence cannot diagnose congestion losses accurately. Incidentally, based on a very different type of experiment, Bolot [6] has observed that congestion losses appear to be random. We believe that our experiments [3] support Bolot s observation, and provide additional insight into packet losses due to congestion and wireless errors. We must also note that the three congestion avoidance techniques were not designed as loss predictors . These congestion avoidance techniques were designed to let the sender ....

S. Biaz and N. Vaidya. Using end-to-end statistics to distinguish congestion and corruption losses : A negative result. Technical Report (draft version), CS Dept., Texas A&M University, Aug. 1997.

....results for different predictors may be compared with each other. 4.3 Experimental Results As results for both the destination hosts are similar in nature, for brevity, here we present results only for destination all purpose gunk.near.net. Results for daedalus.crosslink.net can be found in [4]. Figure 1 plots the measured values of FCP and AP , for the NDG, NTG and V egas predictors. Part (a) plots results for host1 and part (b) for host2. In each graph, six curves are drawn, one for frequency of congestion prediction FCP and another for accuracy AP for each predictor (NDG, NTG and V ....

S. Biaz and N. Vaidya, "Using end-to-end statistics to distinguish congestion and corruption losses : A negative result," Tech. Rep. (draft version), CS Dept., Texas A&M University, Aug. 1997.

No context found.

S. Biaz and N. Vaidya. Using end-to-end statistics to distinguish congestion and corruption losses: A negative result. Technical Report TR97-009, Texas A&M University, August 9, 1997.

No context found.

S. Biaz and N. H. Vaidya; Using End-to-End Statistics to Distinguish Congestion and Corruption Losses: A Negative Result; Technical Report 97- 009, Computer Science, Texas A&M Univ., August 1997.

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