From: Chris Lewis \(chrlewis\) (chrlewis@cisco.com)
Date: Mon Sep 12 2005 - 11:08:57 GMT-3
Tim, I think you've got it right.
aA more interesting case is to consider the combination of priority and
police in MQC, which is actually in use in several large service
providers that offer voice transport as part of their layer 3 VPN
offering.
If you configure just the priority keyword, and there is no congestion
on the link, voice (assuming that is the only traffic you allow in to
the priority queue) can transmit at above the configured rate.
If you configure both priority and police, even if there is no
congestion on the link, voice cannot transmit at above the configured
rate. This is referred to as an "always on" policer, rather than the
congestion aware policer that is part of the priority command.
Why is the always on policer desireable?
Well it comes down to how capacity is planned for in these networks.
Capacity is planned for on a per class basis through the core. The most
expensive class to provision is the voice class, as that class is always
given more bandwidth than goes in to it under normal conditions to
ensure there is no congestion for it even during r-routes and so forth.
Therefore the providers want to make sure their voice classes do not get
over-subscvribed and deliver downgraded performance, so as well as the
priority command, they also configure an explicit policer for the voice
class.
Chris
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Tim
Sent: Monday, September 12, 2005 8:54 AM
To: 'Stefan Grey'; jenseike@start.no; ccielab@groupstudy.com
Subject: RE: SV: Qos question (bandwidth,police).
Stefan,
Have a look at this link:
http://www.cisco.com/warp/public/105/priorityvsbw.html
Based on my read of this doc, it seems more and more to me that having
both the bandwidth and the priority command for the same class of
traffic doesn't make sense.
My reasoning is as follows:
During periods of congestion, the bandwidth and priority commands have
different and mutually exclusive behaviors. Based on this doc, the
bandwidth command allows traffic to exceed the allocated rate but the
priority command does not. Since both can't be true at the same time
for the same traffic, it seems to me both commands shouldn't not be
configured for the same traffic class.
Another difference is that the priority command is used for the purpose
of reducing latency and jitter. It fulfills this function by having
priority access to the hardware queue which is a fifo queue. This doc
isn't clear on exactly what that means but I take it to mean that if
there's any traffic in the priority queue, it will get serviced first
(moved to the h/w queue) up to point of the specified rate. Unlike the
legacy priority-list and priority-group feature, the MQC priority
command won't starve the other queue's.
I'd like to hear from others and either have my thinking confirmed or
corrected.
HTH, Tim
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Stefan Grey
Sent: Sunday, September 11, 2005 10:36 PM
To: jenseike@start.no; ccielab@groupstudy.com
Subject: RE: SV: Qos question (bandwidth,police).
Hi,
Thank you very much for your reply and for the link. But there is
explained just how the police and bandwidth commands are used separetely
but not together. I just need to find the difference between:
1.
policy-map LIMIT
class web
police 100000 conform-action transmit exceed action drop.
and
2.
policy-map LIMIT
class web
bandwidth 100
police 100000 conform-action transmit exceed action drop.
Could anybody tell something?? Because I don't see any. As I think
during the congestions both configs don't let traffic to be more than
100 M and when there is no congestion there are no limiting in both
case. Thanks.
>From: "Jens Petter Eikeland" <jenseike@start.no>
>Reply-To: "Jens Petter Eikeland" <jenseike@start.no>
>To: "'Stefan Grey'" <examplebrain@hotmail.com>,
><ccielab@groupstudy.com>
>Subject: SV: Qos question (bandwidth,police).
>Date: Sun, 11 Sep 2005 23:28:21 +0200
>
>Hi ,
>
>I am sure you will get a better understanding of this two technologies
>by reading findings yourselfe
>
>Here is a some good documents covering both technologies very vel
>
>http://www.cisco.com/en/US/tech/tk543/tk545/tech_design_guides_list.htm
>l
>
>Answer to your question is to use the policing command and conform it
>to 100 meg as in your example
>
>Jens Petter
>
>
>
>-----Opprinnelig melding-----
>Fra: nobody@groupstudy.com [mailto:nobody@groupstudy.com] Pe vegne av
>Stefan Grey
>Sendt: 11. september 2005 23:02
>Til: ccielab@groupstudy.com
>Emne: Qos question (bandwidth,police).
>
>It is needed to limit the input of some WEB traffic to 100 Megabyte.
>
>I saw such config in the book:
>policy-map LIMIT
>class web
>bandwidth 100
>police 100000 conform-action transmit exceed action drop.
>
>Whe could the bandwidth here be used. Whey to command are used together
>here.
>
>As I know police is really limiting traffic to 100 Mb. But bandwidth is
>for custom queuing.
>Could anybody comment this configuration???? What would be the
>difference between this config with or without bandwidth command.
>
>Thanks.
>
>_________________________________________________________________
>MSN Search Ireland has launched - test-drive it today!
>http://search.msn.ie
>
>_______________________________________________________________________
>Subscription information may be found at:
>http://www.groupstudy.com/list/CCIELab.html
>
>_______________________________________________________________________
>Subscription information may be found at:
>http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Sun Oct 02 2005 - 14:40:14 GMT-3