[iccrg] performance comparison on high-speed tcp protocols

[Yong Xia]

Hi,

We have conducted extensive ns2 simulations to evaluate some of the recently
proposed high-speed TCP protocols. The case we consider is that, in the long
run, we have the freedom to deploy AQM and ECN, which has better performance
(particularly, end-to-end latency, which is increasingly important to the
Internet with more and more VoIP traffic) than otherwise. (A great amount of
similar work has also been done, esp. without using AQM in the bottleneck
routers.)

We simulated ten protocols, i.e., SACK, Reno, HSTCP, HTCP, STCP, FAST, BIC,
CUBIC, XCP and VCP. For the first eight protocols, we run RED with ECN
enabled in the bottlenecks. The latter two protocols have their own router
algorithms, so they do not use RED. The details can be seen in our technical
report http://networks.ecse.rpi.edu/~xiay/pub/vcp_tr.ps. More information is
available at http://networks.ecse.rpi.edu/~xiay/vcp.html.

We welcome your comments. Thanks.

Yong
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://oakham.cs.ucl.ac.uk/pipermail/iccrg/attachments/20070308/dd229117/attachment.html