A study of TCP behavior over UBR and ABR
The ability of TCP to provide its own congestion control mechanism allows it to effectively overcome segment loss and avoid congestion collapse. When running TCP over UBR with EPD, we obtain better throughput than with UBR alone. A UBR+EPD based sub-network delivers just as efficient throughput and...
Main Author: | |
---|---|
Format: | Others |
Published: |
1999
|
Online Access: | http://spectrum.library.concordia.ca/874/1/MQ43665.pdf Birdi, Tarlok <http://spectrum.library.concordia.ca/view/creators/Birdi=3ATarlok=3A=3A.html> (1999) A study of TCP behavior over UBR and ABR. Other thesis, Concordia University. |
id |
ndltd-LACETR-oai-collectionscanada.gc.ca-QMG.874 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-LACETR-oai-collectionscanada.gc.ca-QMG.8742013-10-22T03:41:03Z A study of TCP behavior over UBR and ABR Birdi, Tarlok The ability of TCP to provide its own congestion control mechanism allows it to effectively overcome segment loss and avoid congestion collapse. When running TCP over UBR with EPD, we obtain better throughput than with UBR alone. A UBR+EPD based sub-network delivers just as efficient throughput and fair allocation as an ABR+EPD based sub-network. In fact, the Cost-Benefit Ratio, which is an indicator for the packet loss to throughput efficiency, is similar in both scenarios. However, this is assuming that all TCP sources have the same Round Trip Time (RTT). When different RTT sources are implemented, the ABR+EPD sub-network provides better fairness. The final decision deals with enduring the cost of controlling the buffering scheme at the ATM network edge devices, or relying on whatever policy is in place on the ATM switch, albeit at a lower implementation cost. 1999 Thesis NonPeerReviewed application/pdf http://spectrum.library.concordia.ca/874/1/MQ43665.pdf Birdi, Tarlok <http://spectrum.library.concordia.ca/view/creators/Birdi=3ATarlok=3A=3A.html> (1999) A study of TCP behavior over UBR and ABR. Other thesis, Concordia University. http://spectrum.library.concordia.ca/874/ |
collection |
NDLTD |
format |
Others
|
sources |
NDLTD |
description |
The ability of TCP to provide its own congestion control mechanism allows it to effectively overcome segment loss and avoid congestion collapse. When running TCP over UBR with EPD, we obtain better throughput than with UBR alone. A UBR+EPD based sub-network delivers just as efficient throughput and fair allocation as an ABR+EPD based sub-network. In fact, the Cost-Benefit Ratio, which is an indicator for the packet loss to throughput efficiency, is similar in both scenarios. However, this is assuming that all TCP sources have the same Round Trip Time (RTT). When different RTT sources are implemented, the ABR+EPD sub-network provides better fairness. The final decision deals with enduring the cost of controlling the buffering scheme at the ATM network edge devices, or relying on whatever policy is in place on the ATM switch, albeit at a lower implementation cost. |
author |
Birdi, Tarlok |
spellingShingle |
Birdi, Tarlok A study of TCP behavior over UBR and ABR |
author_facet |
Birdi, Tarlok |
author_sort |
Birdi, Tarlok |
title |
A study of TCP behavior over UBR and ABR |
title_short |
A study of TCP behavior over UBR and ABR |
title_full |
A study of TCP behavior over UBR and ABR |
title_fullStr |
A study of TCP behavior over UBR and ABR |
title_full_unstemmed |
A study of TCP behavior over UBR and ABR |
title_sort |
study of tcp behavior over ubr and abr |
publishDate |
1999 |
url |
http://spectrum.library.concordia.ca/874/1/MQ43665.pdf Birdi, Tarlok <http://spectrum.library.concordia.ca/view/creators/Birdi=3ATarlok=3A=3A.html> (1999) A study of TCP behavior over UBR and ABR. Other thesis, Concordia University. |
work_keys_str_mv |
AT birditarlok astudyoftcpbehavioroverubrandabr AT birditarlok studyoftcpbehavioroverubrandabr |
_version_ |
1716605504910786560 |