> This appears to be two parameters: a) delayacks=1 or 3 on the "client
> side" and

According to the docs, delayacks is only for the SYN and FIN packets. May
work well for small transactions, but for ftp should be almost invisible.


--
Doing AIX support was the most monty-pythonesque
activity available at the time.
Eagerly awaiting my thin chocolat mint.

"Jose Pina Coelho" <eresquigal@netcabo.pt> wrote in message
news:Xns96F8CE782848Ceresquigal@194.65.14.158...
> > This appears to be two parameters: a) delayacks=1 or 3 on the "client
> > side" and
>
> According to the docs, delayacks is only for the SYN and FIN packets. May
> work well for small transactions, but for ftp should be almost invisible.

Nope, the SYN is used, in this case, to confirm the sequence no.s
sent/received of the "window" of TCP packets sent over the TCP connection
between Server and Client, this is the "high-level" send/receive window
control mechanism's "tuning knob" for reducing unnecessary protocol overhead
for sending ACKs from Client to Server for a very large and long
transmission from Server to Client, at very high speed.
--
Regards,
Tim Clarke (a.k.a. WBST)

Hi All,

I have finally found a solution for data transfer via this MPLS
network.

What I tried is a open source software named UFTP, which is a multicast
file transfer program, utilizing a protocol based on Starburst MFTP.
UFTP uses UDP instead of TCP.
http://www.tcnj.edu/~bush/uftp.html

My best test results using this UFTP is 36Mbps, or transferring 1GB
file in 4 minutes, comparing the best TCP throughput of 5Mbps.

It seems too much trouble for TCP in this high latency environment.
After some research on high bandwidth/latency enviroment, I found many
companies/institutes use their own protocol to get the best
performance, instead of standard TCP protocol. The typical usage are:
1. Parallel FTP.
http://www.llnl.gov/asci/discom/sc2000_king.html
2. TCP Over UDP
http://www.csm.ornl.gov/~dunigan/net100/atou.html
3. Mutlicast FTP.
http://www.tcnj.edu/~bush/uftp.html

There also is an article for Improving TCP Performance in MPLS but it
seems to difficult to me:
http://networks.cnu.ac.kr/publication/jcn.pdf

Thank you for your help!

Hunter Zhou