Re: Two Rate Policer
From: gladston@br.ibm.com
Date: Thu Oct 21 2004 - 18:21:07 GMT-3
This is a multipart message in MIME format.
--=_alternative 006A4CC283256F34_=
Content-Type: text/plain; charset="US-ASCII"
Thanks you Guys,
I wish you wrote the CD Doc (life would be easier :) )
========
quoted
The rate of filling of the first bucket is PIR, and its size is Be.
========
I am sure it was a typo and you mean the second bucket.
Carlos G Mendioroz <tron@huapi.ba.ar>
21/10/2004 08:51
To
Bob Sinclair <bsinclair@netmasterclass.net>
cc
Alaerte Gladston Vidali/Brazil/IBM@IBMBR, ccielab@groupstudy.com
Subject
Re: Two Rate Policer
To be direct on Gladston question, yes.
Two token and two rate are different things.
To understand the difference you have to learn what the (token) buckets
are used for.
In one bucket alg, there is one bucket which is filled at certain rate
(bps, Bps, pps or the unit at hand) and that has some capacity (in bits,
bytes or packets).
To be able to transmit, you have to take enough "credits" from the
bucket. The problem is that if you don't transmit, then you can
accumulate only the bucket's capacity. Once the bucket is full, overflow
credits are lost.
And if you want to transmit a lot, you can burst as much as capacity at
once but then have to wait for more credits that arrive at fill rate.
Fill rate is CIR. Bucket capacity is Bc (CBS in RFC parlance). The
algorithm is also dubbed two color, because at a given packet, it can
have two outputs: ok/not ok. (conform/exceed)
Then you have two ways of extending this with another bucket:
RFC2697 has another bucket to get credits that overflow the first bucket
up to a certain extra capacity. Then, if you want to transmit and first
bucket credits are not enough, you can take from the second bucket.
If that's ok, then you have "exceed" action, if that's not enough, you
have "violate". Thus you have a three color marker, and still a single
rate. The second bucket size is Be (EBS in RFC parlance).
RFC2698 makes it differently. Instead of having a second bucket to get
the overflown credits, you put another bucket similar to the first in
front of it, with a new rate and a new capacity. This gives you another
marker that is used up front and if you don't conform, you get the
violate action. But if you do, you go to the second (original) bucket
and now you get either conform or exceed.
Note that a conforming packet uses credits from both buckets!
The rate of filling of the first bucket is PIR, and its size is Be.
To complete (and complicate) the the thing, "filling" operations are
synchronous and happen at regular intervals. That's Tc. This is not
strictly conformant with RFC, which asks for CIR times per second
updates of size 1, but it fits the frame relay way.
HTH
Bob Sinclair wrote:
> Gladston,
>
> The major difference between the single-rate and two-rate policer is in
> how tokens get into a second bucket. In the single-rate policer,
> tokens are placed into the second (Be) bucket only when you are below
> the the average rate. It acts like a savings account to let you burst
> above cir. With the two-rate policer the buckets are filled
> independently, you do not have to go below cir in order to put tokens
> in the second bucket.
>
> By analogy, the single rate policer is like getting a $10/week
> allowance. The only way you can spend $15 in a week is to save $5 from a
> previous week. The two rate policer is like getting two allowances per
> week - $10 from your parents and $5 from your grand parents. You can
> spend $15/week without having to first save.
>
> If you want a three-color marker (conform, exceed, violate), you can use
> either method. The RFCs (2697 and 2698) say that the single-rate is
> most useful when the it is the LENGTH of a burst, not its peak rate that
> determines service elilbility. The two-rate is most useful when peak
> rate needs to be enforced separately from a committed rate. The two-rate
> method allows for a sustained excess rate, and seems easier to
> configure, to me, but it really depends on the traffic profile you want
> to enforce.
>
> HTH,
>
> Bob Sinclair
> CCIE #10427, CISSP, MCSE
> www.netmasterclass.net
>
> ----- Original Message ----- From: <gladston@br.ibm.com>
> To: <ccielab@groupstudy.com>
> Sent: Wednesday, October 20, 2004 4:44 PM
> Subject: Two Rate Policer
>
>
>> Kind of confuse with this statement:
>>
>> =================
>> quoted
>> this feature was available, you could police traffic with the
>> single-rate Traffic Policing feature.
>>
http://www.cisco.com/univercd/cc/td/doc/product/software/ios122/122newft/122t/122t4/ft2rtplc.htm
>>
>> ================
>>
>> Because this another Cisco URL seems to says it differently:
>> ================
>> quoted
>> There are currently two types of token bucket algorithms: a single
>> token bucket algorithm and a two token bucket algorithm. A single
>> token bucket system is used when the violate-action option is not
>> specified, and a two token bucket system is used when the
>> violate-action option is specified.
>>
http://www.cisco.com/univercd/cc/td/doc/product/software/ios122/122newft/122t/122t2/ftpoli.htm
>>
>> =================
>>
>>
>> Am I taking an erroneous assumption that "single-rate" and single
>> token" are the same terms?
>>
>> Behind that question is that it is not clear to me what task would
>> require using one or another.
>> police cir pir
>> or
>> police (bps) (burst-normal)...
>>
>> _______________________________________________________________________
>> 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
>
--
Carlos G Mendioroz <tron@huapi.ba.ar> LW7 EQI Argentina
--=_alternative 006A4CC283256F34_=
Content-Type: text/html; charset="US-ASCII"
<br><font size=2 face="sans-serif">Thanks you Guys,</font>
<br>
<br><font size=2 face="sans-serif">I wish you wrote the CD Doc (life would
be easier :) )</font>
<br>
<br><font size=2 face="sans-serif">========</font>
<br><font size=2 face="sans-serif">quoted</font>
<br><font size=2><tt>The rate of filling of the first bucket is PIR, and
its size is Be.</tt></font>
<br><font size=2 face="sans-serif">========</font>
<br>
<br><font size=2 face="sans-serif">I am sure it was a typo and you mean
the second bucket.</font>
<br>
<br><font size=2 face="sans-serif"><br>
</font>
<br>
<br>
<br>
<table width=100%>
<tr valign=top>
<td width=40%><font size=1 face="sans-serif"><b>Carlos G Mendioroz <tron@huapi.ba.ar></b>
</font>
<p><font size=1 face="sans-serif">21/10/2004 08:51</font>
<td width=59%>
<table width=100%>
<tr>
<td>
<div align=right><font size=1 face="sans-serif">To</font></div>
<td valign=top><font size=1 face="sans-serif">Bob Sinclair <bsinclair@netmasterclass.net></font>
<tr>
<td>
<div align=right><font size=1 face="sans-serif">cc</font></div>
<td valign=top><font size=1 face="sans-serif">Alaerte Gladston Vidali/Brazil/IBM@IBMBR,
ccielab@groupstudy.com</font>
<tr>
<td>
<div align=right><font size=1 face="sans-serif">Subject</font></div>
<td valign=top><font size=1 face="sans-serif">Re: Two Rate Policer</font></table>
<br>
<table>
<tr valign=top>
<td>
<td></table>
<br></table>
<br>
<br>
<br><font size=2><tt>To be direct on Gladston question, yes.<br>
Two token and two rate are different things.<br>
</tt></font>
<br><font size=2><tt>To understand the difference you have to learn what
the (token) buckets<br>
are used for.<br>
</tt></font>
<br><font size=2><tt>In one bucket alg, there is one bucket which is filled
at certain rate<br>
(bps, Bps, pps or the unit at hand) and that has some capacity (in bits,<br>
bytes or packets).<br>
To be able to transmit, you have to take enough "credits" from
the<br>
bucket. The problem is that if you don't transmit, then you can<br>
accumulate only the bucket's capacity. Once the bucket is full, overflow<br>
credits are lost.<br>
And if you want to transmit a lot, you can burst as much as capacity at<br>
once but then have to wait for more credits that arrive at fill rate.<br>
</tt></font>
<br><font size=2><tt>Fill rate is CIR. Bucket capacity is Bc (CBS in RFC
parlance). The<br>
algorithm is also dubbed two color, because at a given packet, it can<br>
have two outputs: ok/not ok. (conform/exceed)<br>
</tt></font>
<br><font size=2><tt>Then you have two ways of extending this with another
bucket:<br>
RFC2697 has another bucket to get credits that overflow the first bucket<br>
up to a certain extra capacity. Then, if you want to transmit and first<br>
bucket credits are not enough, you can take from the second bucket.<br>
If that's ok, then you have "exceed" action, if that's not enough,
you<br>
have "violate". Thus you have a three color marker, and still
a single<br>
rate. The second bucket size is Be (EBS in RFC parlance).<br>
</tt></font>
<br><font size=2><tt>RFC2698 makes it differently. Instead of having a
second bucket to get<br>
the overflown credits, you put another bucket similar to the first in<br>
front of it, with a new rate and a new capacity. This gives you another<br>
marker that is used up front and if you don't conform, you get the<br>
violate action. But if you do, you go to the second (original) bucket<br>
and now you get either conform or exceed.<br>
Note that a conforming packet uses credits from both buckets!<br>
The rate of filling of the first bucket is PIR, and its size is Be.<br>
</tt></font>
<br><font size=2><tt>To complete (and complicate) the the thing, "filling"
operations are<br>
synchronous and happen at regular intervals. That's Tc. This is not<br>
strictly conformant with RFC, which asks for CIR times per second<br>
updates of size 1, but it fits the frame relay way.<br>
</tt></font>
<br><font size=2><tt>HTH<br>
</tt></font>
<br>
<br><font size=2><tt>Bob Sinclair wrote:<br>
> Gladston,<br>
><br>
> The major difference between the single-rate and two-rate policer
is in<br>
> how tokens get into a second bucket. In the single-rate policer,<br>
> tokens are placed into the second (Be) bucket only when you are below<br>
> the the average rate. It acts like a savings account to let
you burst<br>
> above cir. With the two-rate policer the buckets are filled<br>
> independently, you do not have to go below cir in order to put
tokens<br>
> in the second bucket.<br>
><br>
> By analogy, the single rate policer is like getting a $10/week<br>
> allowance. The only way you can spend $15 in a week is to save $5
from a<br>
> previous week. The two rate policer is like getting two allowances
per<br>
> week - $10 from your parents and $5 from your grand parents. You
can<br>
> spend $15/week without having to first save.<br>
><br>
> If you want a three-color marker (conform, exceed, violate), you can
use<br>
> either method. The RFCs (2697 and 2698) say that the single-rate
is<br>
> most useful when the it is the LENGTH of a burst, not its peak rate
that<br>
> determines service elilbility. The two-rate is most useful when
peak<br>
> rate needs to be enforced separately from a committed rate. The two-rate<br>
> method allows for a sustained excess rate, and seems easier to<br>
> configure, to me, but it really depends on the traffic profile you
want<br>
> to enforce.<br>
><br>
> HTH,<br>
><br>
> Bob Sinclair<br>
> CCIE #10427, CISSP, MCSE<br>
> www.netmasterclass.net<br>
><br>
> ----- Original Message ----- From: <gladston@br.ibm.com><br>
> To: <ccielab@groupstudy.com><br>
> Sent: Wednesday, October 20, 2004 4:44 PM<br>
> Subject: Two Rate Policer<br>
><br>
><br>
>> Kind of confuse with this statement:<br>
>><br>
>> =================<br>
>> quoted<br>
>> this feature was available, you could police traffic with the<br>
>> single-rate Traffic Policing feature.<br>
>> http://www.cisco.com/univercd/cc/td/doc/product/software/ios122/122newft/122t/122t4/ft2rtplc.htm>
>><br>
>> ================<br>
>><br>
>> Because this another Cisco URL seems to says it differently:<br>
>> ================<br>
>> quoted<br>
>> There are currently two types of token bucket algorithms: a single<br>
>> token bucket algorithm and a two token bucket algorithm. A single<br>
>> token bucket system is used when the violate-action option is
not<br>
>> specified, and a two token bucket system is used when the<br>
>> violate-action option is specified.<br>
>> http://www.cisco.com/univercd/cc/td/doc/product/software/ios122/122newft/122t/122t2/ftpoli.htm>
>><br>
>> =================<br>
>><br>
>><br>
>> Am I taking an erroneous assumption that "single-rate"
and single<br>
>> token" are the same terms?<br>
>><br>
>> Behind that question is that it is not clear to me what task would<br>
>> require using one or another.<br>
>> police cir pir<br>
>> or<br>
>> police (bps) (burst-normal)...<br>
>><br>
>> _______________________________________________________________________<br>
>> Subscription information may be found at:<br>
>> http://www.groupstudy.com/list/CCIELab.html>
><br>
><br>
> _______________________________________________________________________<br>
> Subscription information may be found at:<br>
> http://www.groupstudy.com/list/CCIELab.html>
><br>
</tt></font>
<br><font size=2><tt>--<br>
Carlos G Mendioroz <tron@huapi.ba.ar> LW7 EQI Argentina</tt></font>
<br>
<br>
--=_alternative 006A4CC283256F34_=--
This archive was generated by hypermail 2.1.4
: Sat Nov 06 2004 - 17:11:51 GMT-3