[c-nsp] Multilink PPP throughput instable

[Rodney Dunn]

Try putting WRED on the bundle and see if that helps.

I wonder if you are seeing TCP backoff kicking in when
the links get congested.

On Wed, Apr 27, 2005 at 12:04:40AM +0800, tony kam wrote:
> We did the test and the pair of servers were able to flood the individual E1 
> link up to port speed very stable with upload and download ftp at the same 
> time. Once both E1 links were bonded together by MLPPP, there was throughput 
> variation from 3.2 to 3.9Mbps with the same ftp upload and download at the 
> same time.
> 
> Please note that hardware load-balancing like IMUX or IMA, without such 
> problem.  This is the reason why I suspect MLPPP, a software-based load 
> balancing, cause the problem.
> 
> If possible, I would like to hear comments from Cisco folks as well.
> 
> Thanks.
> 
> ----- Original Message ----- 
> From: "Mikael Abrahamsson" <swmike at swm.pp.se>
> To: "tony kam" <chiwaikam at hotmail.com>
> Cc: <cisco-nsp at puck.nether.net>
> Sent: Tuesday, April 26, 2005 2:34 AM
> Subject: Re: [c-nsp] Multilink PPP throughput instable
> 
> 
> > On Mon, 25 Apr 2005, tony kam wrote:
> >
> >> Thanks Mikael! If it is still getting the same result after all these TCP 
> >> optimization, do you think there is something to do with MLPPP 
> >> configurations? Or it is the behavior of MLPPP bundle under heavy 
> >> congestion?
> >> Please note that individual E1 link is working perfect in bidirectional 
> >> ftp, i.e., put and get at the same time on the same pair of servers.
> >
> > Well, you could try and see if you get the same problem without MLPPP. My 
> > guess is that it's a generic problem and not MLPPP related.
> >
> > -- 
> > Mikael Abrahamsson    email: swmike at swm.pp.se
> > 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/