Strange behavior in CBQ for Reno TCP traffic
Thu, 19 Aug 1999 18:00:06 +0530 (IST)
On Thu, 19 Aug 1999, Chetan Kumar wrote:
I found that this problem vanishes, when I have audio, video and
ftp have the same priority. So there is something to do with priority, is
this a bug in CBQ implementation in NS ???
*> I was running couple of CBQ simulations, this is the senirio
*>1) A flat tree structure with classes
*>Data 62,2 source FTP over TCP/Reno
*>Audio 3,1 source CBR size 190, at 1 ms interval
*>Video 32,1 source CBE size 500, at 2 ms interval
*>This is same as cbqWRR test case in the validation test cases, expect for
*>FTP traffic replacing data (CBR) traffic.
*>Now with one ftp session the simulation looks OK, with 2 ftp session the
*>FTP traffic is more flucating, and with 3 ftp session the total ftp
*>throughput reduces to almost zero at around 30 second.
*>Has anybody experienced this kind of behavior, is there any pitfall in
*>what I am doing.