From: CCIE FUN (ccieexam2002@yahoo.com)
Date: Fri Mar 28 2003 - 13:34:05 GMT-3
Scott
Based on what you are saying couldn't this approach be
taken:
!
class-map match-all TELNET
match protocol telnet
class-map match-all DLSW
match protocol dlsw
class-map match-all FTP
match protocol ftp
class-map match-all DNS
match protocol dns
!
policy-map MORE-FUN
class TELNET
bandwidth percent 9
queue-limit 100
class DNS
bandwidth percent 13
queue-limit 100
class FTP
bandwidth percent 26
queue-limit 20
class DLSW
bandwidth percent 43
queue-limit 20
class class-default
bandwidth percent 9
queue-limit 20
!
--- "Scott M. Livingston" <scottl@sprinthosting.net>
wrote:
> Fabrice,
>
> I am still debating between the two formulas
> tonight.
>
> What do you think about this? Referencing CQ; I am
> sure you know the
> formula in how to calculate BW distribution in bytes
> for queues based on
> packet size of protocol.... If that was wordy and
> confusing here is a
> url that illustrates;
>
http://www.cisco.com/univercd/cc/td/doc/product/software/ios121/121cgcr/
> qos_c/qcprt2/qcdconmg.htm#1001366
>
> So then, when we use this Cisco formula to determine
> bytes for each
> queue we are not referencing the 25% default BW for
> system traffic. We
> are configuring for 100% of the link, but in reality
> the router will
> only allow us 75% of the interface; much in the same
> way my original
> formula was doing when converting from CQ to CBWFQ?
>
>
> What I mean is the following - when we are just
> configuring CQ and
> determining the byte size of each queue based off
> protocol packet size
> we are aware there is that implicit 25% default
> queue that does not come
> into play w/ our formula. Also, when we convert our
> CQ bytes to a CBWFQ
> bandwidth my original formula does not take into
> account that same 25%
> "system queue" (ya nobody calls it that, but I am
> just trying to make a
> point).
>
> So, by not referencing 75% anywhere in our equation
> we are basically
> comparing apples to apples when converting the byte
> count (BW) in CQ to
> bandwidth in CBWFQ? Does that make sense? Am I
> missing something?
>
> I probably didn't do a very good job explaining that
> so let me know if I
> was too confusing.
>
> Thanks,
> scott
>
> -----Original Message-----
> From: nobody@groupstudy.com
> [mailto:nobody@groupstudy.com] On Behalf Of
> Scott M. Livingston
> Sent: Thursday, March 27, 2003 4:20 PM
> To: 'Fabrice Bobes'; 'Abdul Waheed Ghaffar';
> ccielab@groupstudy.com
> Subject: RE: QOS
>
> MMMMMM. You diggin' deep on this one. I understand
> what you are
> explaining now; I forgot about that system queue.
> HMMMM I will have to
> look at this a little harder, but your argument
> sounds pretty good to me
> right now.
>
> Any contention from anyone or does everyone see it
> like Fabrice does?
>
> Thanks Fabrice!
> scott
>
> -----Original Message-----
> From: Fabrice Bobes [mailto:study@6colabs.com]
> Sent: Thursday, March 27, 2003 4:07 PM
> To: 'Scott M. Livingston'; 'Abdul Waheed Ghaffar';
> ccielab@groupstudy.com
> Subject: RE: QOS
>
> Well, you are still assigning the same percentage of
> bandwidth
> relatively to the max allocable bandwidth (75%). Not
> sure it's very
> clear.
>
> Also, with CQ, there is this queue #0 (system queue)
> that you don't
> configure. Queue #0 borrows the bandwidth that it
> needs. When we say
> that queue #1 in CQ takes 9% of the total bandwidth,
> it's not completely
> true. It takes 9% of the bandwidth that we can
> assign to the 16
> configurable queues. The queue 0 shouldn't take 25%
> of the bandwidth but
> like CBWFQ, there is this idea of a max bandwidth
> that you allocate to
> the queues and it's inferior to 100% of the
> bandwidth of the link.
>
> Fabrice
>
>
> -----Original Message-----
> From: Scott M. Livingston
> [mailto:scottl@sprinthosting.net]
> Sent: Thursday, March 27, 2003 1:37 PM
> To: 'Fabrice Bobes'; 'Abdul Waheed Ghaffar';
> ccielab@groupstudy.com
> Subject: RE: QOS
>
> Hmmmm? Interesting thinking on your part.
>
> If you do it that way though you really wouldn't be
> allocating the same
> amount of BW that the CQ was allocating to that
> traffic right? Am I
> missing something here?
>
> Scott
>
> -----Original Message-----
> From: Fabrice Bobes [mailto:study@6colabs.com]
> Sent: Thursday, March 27, 2003 3:21 PM
> To: 'Scott M. Livingston'; 'Abdul Waheed Ghaffar';
> ccielab@groupstudy.com
> Subject: RE: QOS
>
> Hi Scott,
>
> If I may, I wouldn't allocate 100% of the bandwidth
> to the classes and
> would keep the default to 75 %. If you want to
> assign more than 75% to
> the classes, that's fine but not up to 100%.
>
> In other words, I'd rather use this formula for
> queue #1:
> 128 kbps * (1000/11500) * 0.75 = 8 (bandwidth 8)
> In percentage, this gives us:
> (1000/11500) * 0.75 = 6.5 (bandwidth percent 6)
>
> May be I am missing something here so any input?
>
> Fabrice
> http://www.6colabs.com
>
>
> -----Original Message-----
> From: nobody@groupstudy.com
> [mailto:nobody@groupstudy.com] On Behalf Of
> Scott M. Livingston
> Sent: Monday, March 24, 2003 8:01 PM
> To: 'Abdul Waheed Ghaffar'; ccielab@groupstudy.com
> Subject: RE: QOS
>
> Here is my attempt. For these scenarios if I can
> spot the bandwidth
> percent real quick then I just use the 'bandwidth
> percent' command. In
> my own freakish way of thinking I can convert it
> quicker using the
> formula below. Tell me what you all think and please
> let me know if I
> messed up somewhere.
>
> BW of Link in kbps * (byte count / total byte count)
> = BW
>
> For example:
>
> Queue #1
> ---------
> 128 kbps * (1000 / 11500) = 11
>
> !
> class-map match-all TELNET
> match protocol telnet
> class-map match-all DLSW
> match protocol dlsw
> class-map match-all FTP
> match protocol ftp
> class-map match-all DNS
> match protocol dns
> !
> policy-map MORE-FUN
> class TELNET
> bandwidth 11
> queue-limit 100
> class DNS
> bandwidth 16
> queue-limit 100
> class FTP
> bandwidth 33
> queue-limit 20
> class DLSW
>
=== message truncated ===
This archive was generated by hypermail 2.1.4 : Sat Apr 05 2003 - 08:51:46 GMT-3