From: Salau, Yemi (yemi.salau@siemens.com)
Date: Tue May 20 2008 - 06:15:05 ART
Hi Guys,
This is my own 2 kobos worth's:
I know many have different takes on this topic, ie. cpu-processed or
not, transit or non-transit, inbound or outbound .... ok now, forget the
control-plane story because Cisco says it only applies to "police rate"
& not "police cir pir"
Forget the inbound/outbound because "police bps" & "police cir" can be
applied the same way. But in the real world, would you rather pay
service charges for your traffic to fly all the way from Miami to Paris?
or would you just police as close to the traffic source as possible, yes
ofcourse, hence why I love doing outbound policing rather than inbound.
Ofcourse I think SP guys do inbound just to make sure their customers
don't do traffic above the Committed Information Rate.
To cut the long story short, "police bps" achieves the same thing as
"pir cir". Actually police cir is the short form of the long command
"police cir pir" which was put in so we can use two rate buckets for our
policing, as opposed to one rate bucket with "police bps" - bps here by
the way refers to average rate, which essentially is our committed rate
right? CIR!
But with "police cir pir" you have the luxury of using 2 buckets Tc & Tp
as Cisco calls them to police your traffic. However, if you use one of
those two buckets say Tc token bucket which gets updated at the CIR
value each time a packet arrives at the two-rate policer, then it will
achieve the similar results as using "police bps"
So, "police bps" = "police cir" but "police cir" is the short form of
"police cir pir" right? which uses 2 token buckets. So "police bps" !=
"police cir pir"
Another thing I noticed is that, If you're using "police bps" and your
packet size exceed the average rate, obviously the exceed action gets
applied until the packet burst into the violation side of things. This
is same if you're using "police cir", but if you're using "police cir
pir"; the packet gets measured with Tp bucket first and applies the
violate action if this exceed the Tp bucket, if your packet is less than
the Tp bucket, then the packet is measured against the Tc bucket, then
exceed action applies, if not, then conform action all the way.
Please, there is a Cisco figure I'll like you to take a look at:
http://www.cisco.com/en/US/docs/ios/qos/command/reference/qos_n1.html#wp
1027089
Hope this helps, please if I don't get one or two things right, please
feel free to underline them ....
Many Thanks
Yemi Salau
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Matt Bentley
Sent: Monday, May 19, 2008 9:51 PM
To: Dale Kling
Cc: Sadiq Yakasai; Cisco certification
Subject: Re: Police Vs. Police CIR
Hi Dale:
I had read over the link, but couldn't make much of it. I found an
archived
thread about this topic (
http://www.groupstudy.com/archives/ccielab/200508/msg01075.html), but it
seems to be a brief overview. Any other help would be greatly
appreciated.
Thanks.
On Mon, May 19, 2008 at 4:30 PM, Dale Kling <dalek77@gmail.com> wrote:
> I'm reading over this as we type, I never really thought about it
until you
> brought it up. Here's a link to the DOCCD I just found on it.
>
>
>
http://www.cisco.com/en/US/docs/ios/12_4/qos/configuration/guide/h2RTplc
.html
>
> regards,
>
> Dale
>
> Let me know what you make of it.
>
> On Mon, May 19, 2008 at 4:24 PM, Matt Bentley <mattdbentley@gmail.com>
> wrote:
>
>> Hi Sadiq:
>>
>> Thanks for the explanation. I understand what you said, but I still
see
>> it
>> - at least from a question perspective to use the policy x y z
command for
>> traffic that enters/exits an interface. Do you mean
hardware-processed
>> traffic would use the police x y z command whereas CPU-processed
traffic
>> uses police cir?
>>
>> Thanks again.
>>
>> On Mon, May 19, 2008 at 4:10 PM, Sadiq Yakasai <sadiqtanko@gmail.com>
>> wrote:
>>
>> > Hi Matt,
>> >
>> > police x y z command polices control plane traffic. This is traffic
>> > that terminates on a router (non transit traffic).
>> >
>> > While police cir x y z polices data plane (transit) traffic that is
>> > either coming in or going out an interface.
>> >
>> > HTH
>> > Sadiq
>>
>>
>>
This archive was generated by hypermail 2.1.4 : Mon Jun 02 2008 - 06:59:17 ART