SCTP Discussions

Text archives Help


Re: [SCTP-Discussion] cwnd grow during congestion avoidance


Chronological Thread 
  • From: Michael Tuexen <Michael.Tuexen AT micmac.franken.de>
  • To: Atanas Matev <razor AT bulhack.org>
  • Cc: discussion AT sctp.de
  • Subject: Re: [SCTP-Discussion] cwnd grow during congestion avoidance
  • Date: Tue May 3 22:12:01 2005
  • List-id: SCTP Discussions <discussion.sctp.de>

Hi Atanas,

thank you very much for reporting this. We'll look into this.

Best regards
Michael

On May 2, 2005, at 4:50 PM, Atanas Matev wrote:

Hello,
I am a student and part of my graduation work touches the congestion
control mechanism of SCTP. I choose "sctplib" as SCTP implementation but
there are some problems which I can not explain.
So I use "echo_tool" and "echo_server" to monitor the values of cwnd and
ssthresh. In the phase of slow start everything looks fine and cwnd in
incremented by 1*MTU (or the size of the outstanding data chunks
acknowledged) per SACK until it reaches the ssthresh value. In theory
slow start ends here and we goes in the congestion avoidance phase. Cwnd
have to continue growing up by 1*MTU per RTT but it doesn't happens:

bash-2.05b# ./echo_tool -v -i -s 192.168.1.2 -d 192.168.1.2 -M -n 10 -p
2 -l 1024 > results.log
.
.
Asoc:1 Path:0 Ch:4 By:4096 rto:1000 srtt:1
qu:0 osb:10400 cwnd:105140 ssthresh:109568
Asoc:1 Path:0 Ch:3 By:3072 rto:1000 srtt:1
qu:0 osb:10400 cwnd:108260 ssthresh:109568
Asoc:1 Path:0 Ch:3 By:3072 rto:1000 srtt:1
qu:0 osb:10400 cwnd:110340 ssthresh:109568
Asoc:1 Path:0 Ch:3 By:3072 rto:1000 srtt:1
qu:0 osb:10400 cwnd:110340 ssthresh:109568
Asoc:1 Path:0 Ch:4 By:4096 rto:1000 srtt:1
qu:0 osb:10400 cwnd:110340 ssthresh:109568
.
.

And cwnd keeps this value forever. Why is this happening? I also create
a simulation model using NS-2 and the results there seems much closer to
the expected ones - when cwnd became greater than ssthresh, it continue
ascending with 1*MTU per RTT.

Is this some kind of disadvantage in "sctplib" or I am doing something
wrong?

Thanks in advice!

Best Regards:
Atanas Matev
_______________________________________________
discussion mailing list
discussion AT sctp.de
http://www.sctp.de/mailman/listinfo/discussion







Archive powered by MHonArc 2.6.19+.

Top of page