Subject: BOUNCE cisco-nsp@qual.net: Non-member submission from [Scott Whyte <swhyte@cisco.com>]
>From owner-cisco-nsp Sun Jul 5 19:00:24 1998
Received: from ce-nfs-1.cisco.com (ce-nfs-1.cisco.com [171.68.201.251]) by nic.iagnet.net (8.8.8/IAG/CICNet) with ESMTP id TAA24077 for <cisco-nsp@qual.net>; Sun, 5 Jul 1998 19:00:24 -0400 (EDT)
Received: from swhyte-pc.cisco.com (wwhyte-isdn.cisco.com [171.70.229.22]) by ce-nfs-1.cisco.com (8.8.5/CA/950118) with SMTP id PAA07284; Sun, 5 Jul 1998 15:59:52 -0700 (PDT)
Message-Id: <3.0.5.32.19980705155939.007e89b0@ce-nfs-1.cisco.com>
X-Sender: wwhyte@ce-nfs-1.cisco.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 3.0.5 (32)
Date: Sun, 05 Jul 1998 15:59:39 -0700
To: cisco-nsp@qual.net
From: Scott Whyte <swhyte@cisco.com>
Subject: Re: [nsp] CAR - does it buffer?
Cc: amb@gxn.net
In-Reply-To: <199807051828.TAA11647@diamond.xara.net>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
At 07:28 PM 7/5/98 +0100, Alex Bligh wrote:
>So am I mistaken about CAR, or is the implementation no more
>than dumb policing? And if so, how do I get decent traffic shaping
>for a web farm?
No, you are correct about CAR, it is only a token bucket. It has two
purposes, setting IP precedence, and rate limiting access bandwidth. You
need queueing to reorder packets for traffic shaping.
If you use CAR to set the IP precedence at ingress, you can then use (say)
dWFQ to utilize those precedence bits in weighting various flows. That
should get you the traffic shaping you desire, as dWFQ controls how much
bandwidth each flow can use during congestion.
These are all 11.1CC features...
---- Scott Whyte |"Computers are different from telephones. Network Supported Accounts | Telephones ring. Computers do not ring." CCIE | -- Andrew S. Tanenbaum
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:13:13 EDT