From: Marco Umer (ccie@umer.it)
Date: Sun Jul 13 2003 - 18:43:16 GMT-3
Hi,
I have seen more detailed and real world working answer, but I think that
30M 15K 15K (as described in the book) is because 15K is the lowest value
you can configure for 30M of CIR: CIR/2000.
If you try less you will see:
down(config-if)#rate-limit output 30000000 10000 10000 conform-action
transmit$
Illegal normal burst size
Increasing normal burst size to 15000
Illegal extended burst size
Increasing extended burst size to 15000
Regards,
Marco
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
Brian McGahan
Sent: Thursday, July 10, 2003 11:27 PM
To: 'Brian McGahan'; 'Jonathan V Hays'; 'Jim Phillipo';
ccielab@groupstudy.com
Subject: QOS CAR example
Group,
As a follow up to my previous post, I did some testing to
illustrate how the burst values of the CAR statement affect the
configured rate.
The following output is from a router connected to the internet
which is policing NNTP traffic (usenet news). I am downloading files
from alt.binaries.sounds.mp3 (for research purposes only of course), and
changing the burst values of the police statement to show the effect.
The goal of our policing is to keep our NNTP traffic at or below 80kpbs.
In the first example, I entered the 'police 80000' statement.
This results in the default values of:
Police 80000 2500 2500 conform-action transmit exceed-action drop
http://www.cyscoexpert.com/adminimages/car4.gif
As you can see from the output, I am averaging a rate of about
30kbps. This is much lower than the expected target rate.
In the next example, the burst values are modified to meet the
recommended formula. The 'police 80000 15000 30000' statement was
entered, which resulted in:
Police 80000 15000 30000 conform-action transmit exceed-action drop
http://www.cyscoexpert.com/adminimages/car3.gif
As you can see from this output, the average rate is in the high
70kbps range, which is very close to our desired target rate.
The calculation for CAR values should therefore be as follows:
Normal_burst bytes = target_rate bps * ( 1 byte / 8 bits ) * 1.5 seconds
Excess_burst bytes = normal_burst * 2
HTH,
Brian McGahan, CCIE #8593
Director of Design and Implementation
brian@cyscoexpert.com
CyscoExpert Corporation
Internetwork Consulting & Training
Toll Free: 866.CyscoXP
Fax: 847.674.2625
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf
Of
> Brian McGahan
> Sent: Thursday, July 10, 2003 2:32 PM
> To: 'Jonathan V Hays'; 'Jim Phillipo'; ccielab@groupstudy.com
> Subject: RE: More QOS CAR example
>
> Jim & Jonathan,
>
> Configuring the burst values to correspond to the formula *IS*
> necessary. Test it with actual traffic and you will see that the rate
> does not correctly conform unless you use the recommended burst
values.
> I'm not sure where these authors are getting their burst values from,
> but they're not correct.
>
> HTH
>
> Brian McGahan, CCIE #8593
> Director of Design and Implementation
> brian@cyscoexpert.com
>
> CyscoExpert Corporation
> Internetwork Consulting & Training
> Toll Free: 866.CyscoXP
> Fax: 847.674.2625
>
>
> > -----Original Message-----
> > From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf
> Of
> > Jonathan V Hays
> > Sent: Wednesday, July 09, 2003 6:37 PM
> > To: 'Jim Phillipo'; ccielab@groupstudy.com
> > Subject: RE: More QOS CAR example
> >
> > > -----Original Message-----
> > > From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On
> > > Behalf Of Jim Phillipo
> > > Sent: Wednesday, July 09, 2003 6:48 PM
> > > To: ccielab@groupstudy.com
> > > Subject: More QOS CAR example
> > >
> > >
> > > This is from VEGESNA's book
> > >
> > > All traffic except HTTP over a rate of 15 Mbps is marked with
> > > IP Prec 4
> > > HTTP over 15 Mbps is marked with 0
> > > This is a 30 Mbps pipe from provider
> > >
> > > There answer:
> > >
> > > rate-limit input 30000000 15000 15000 conform-action continue
> > > exceed-action-drop
> > > rate-limit input access-group 101 15000000 10000 10000
> conform-action
> > > set-prec-transmit 4 exceed-action set-prec-transmit 0
> > > rate-limit input 30000000 15000 15000 conform-action
> > > set-prec-transmit 4
> > > exceed-action set-prec-transmit 4
> > >
> > > access-list 101 permit tcp any any eq www
> > > access-list 101 permit tcp any eq www any <This is for web
> > > traffic leaving
> > > the site correct ?>
> > >
> > > How do they arrive at 15000 and 10000 respectively ?
> >
> > [snip]
> >
> > Good observation.
> >
> >
>
http://www.cisco.com/univercd/cc/td/doc/product/software/ios122/122cgcr/
> > fqos_c/fqcprt4/qcfpolsh.htm#1000977
> >
> > The above reference on the Doc CD gives the CAR formula as:
> > <quote>
> > Recommended Burst Values
> > Cisco recommends the following values for the normal and extended
> burst
> > parameters:
> >
> > normal burst = configured rate * (1 byte)/(8 bits) * 1.5 seconds
> > extended burst = 2 * normal burst
> > </quote>
> >
> > Here are the calculations (commas added for readability - my
apologies
> > to Europeans and others who don't use commas to break up the
thousands
> > ;-)
> >
> > configured rate = 30,000,000 bps
> > normal burst = 30000000/8 * 1.5 = 5,625,000 bytes
> > extended burst = 2 * 5625000 = 11,250,000 bytes
> >
> > configured rate = 10,000,000 bps
> > normal burst = 10000000/8 * 1.5 = 1,875,000 bytes
> > extended burst = 2 * 1875000 = 3,750,000 bytes
> >
> > Vegesna is obviously *not* using this formula. Maybe there is a
> > saturation point or ceiling for burst size? (This is no more than a
> > SWAG.)
> >
> > Perhaps the "experts" will chime in ...
> >
> > Jonathan
> >
> >
> >
>
This archive was generated by hypermail 2.1.4 : Wed Aug 06 2003 - 06:52:38 GMT-3
