RE: [nsp] generic traffic shaping

From: Dave Qi (dqi@bloomberg.com)
Date: Thu May 24 2001 - 18:57:27 EDT


Lee,

CAR can be applied on both inbound and outbound interfaces.

there are other differences b/t CAR and GTS.
check out: http://www.cisco.com/warp/public/cc/pd/iosw/tech/carat_wp.htm

-Dave

-----Original Message-----
From: LEE Jin Hian [mailto:jhlee@starhub.com.sg]
Sent: Wednesday, May 23, 2001 3:40 AM
To: 'Charles Sprickman'
Cc: cisco-nsp@puck.nether.net
Subject: RE: [nsp] generic traffic shaping

Charles,

As I understand it, the differences between CAR & GTS are:

a) GTS
- performed on an output (egress) interface
- involves queueing & buffering to smooth out peaks in traffic
- drops packets only when buffer memory runs out

b) CAR
- performed on an input (ingress) interface
- pure policing, i.e. any traffic that exceeds the burst values is dropped
immediately

AFAIK:

i) GTS should be used by an SP's customers to ensure that they stay within
the bandwidth limits that they have purchased.

ii) SPs will probably run CAR on the ingress interface if required. At
least, we do.

iii) GTS is much preferable to CAR because it isn't so prone to dropping
packets. Hence, it's much less likely to require packet re-transmits and
trigger TCP's congestion control mechanisms.

Regards,
Jin

-----Original Message-----
From: Charles Sprickman [mailto:spork@inch.com]
Sent: 19 May 2001 01:01
To: George Robbins
Cc: gert@greenie.muc.de; goemon@anime.net; cisco-nsp@puck.nether.net
Subject: Re: [nsp] generic traffic shaping

On Fri, 18 May 2001, George Robbins wrote:

> What are you wanting? Whether it's simple or extended, the second
> part is a wildcard mask and .14 is 00001110 which will match only
> eight even IP addresses. This is probably not what you intended.

Yep, as someone else pointed out that should be .15 (255-mask).

I also have another rule with a full C, and seeing similar results.

Anyone use this stuff, or is CAR better? This seemed more "gentle" than
CAR; at least going by how the docs described it...

I still see no affect, and I'm throwing it on a block doing about 1.5 Mb
constant and trying to keep it down to 700K/b.

I'm going to play with CAR now.

Thanks,

C

> George
>
> > Hi,
> >
> > On Thu, May 17, 2001 at 03:23:26PM -0700, Dan Hollis wrote:
> > > On Thu, 17 May 2001, Charles Sprickman wrote:
> > > > access-list 61 permit 192.168.0.0 0.0.0.14
> > >
> > > access-list 61 permit ip 192.168.0.0 0.0.0.15 any
> >
> > As it's a simple access list, no :-) - but maybe that's the problem?
> >
> > gert
> >
> > --
> > USENET is *not* the non-clickable part of WWW!
> >
//www.muc.de/~gert/
> > Gert Doering - Munich, Germany
gert@greenie.muc.de
> > fax: +49-89-35655025
gert.doering@physik.tu-muenchen.de
> >
> >
>



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:38 EDT