SCTP Discussions

Text archives Help


[SCTP-Discussion] Re: Wonder about performance of SCTP


Chronological Thread 
  • From: Michael Tuexen <Michael.Tuexen AT micmac.franken.de>
  • To: "Alfred Boucek" <Ing.AlfredBoucek AT gmx.at>
  • Cc: discussion AT sctp.de
  • Subject: [SCTP-Discussion] Re: Wonder about performance of SCTP
  • Date: Mon Apr 8 09:47:00 2002
  • List-id: SCTP Discussions <discussion.sctp.de>

Alfred,

this prototype implementation has not been optimized for performance. The
focus is more on features and the possibility to have a tool to analyze the
performance the protocol, not the specific implementation.
Please note also that this implementation is a user land implementation
which in some case means that it can not be as fast as a kernel implementation.

We may look into performance optimization when all features have been build
in and have been tested.

You might want to look into the SCTP kernel implementation built for the KAME
stack runnning on *BSD. It is available at www.sctp.org.

Best regards
Michael
On Monday, April 8, 2002, at 08:27 AM, Alfred Boucek wrote:

I try to use SCTP but I wonder about performance I get up to now. I use
SCTPLIB-1.0.0-pre17 and compared to Linux-TCP. SCTP needs round double
amount of CPU time to send data on a 10Base2 LAN. The throughput I can get
with SCTP is in my current test scenario limited by the power of my PC. This
shows up especially on a 100BaseT interface. Is there an easy explanation
for this? Do you have similar results? Are there other implementations more
preferment? How did you measure the throughput of SCTP? I would be glad to
get some figures to compare.

Thank you for our resonse
Alfred Boucek



Michael.Tuexen AT micmac.franken.de




  • [SCTP-Discussion] Re: Wonder about performance of SCTP, Michael Tuexen, 04/08/2002

Archive powered by MHonArc 2.6.19+.

Top of page