Re: QoS default-class - oversubscription

From: Pierre-Alex (paguanel@hotmail.com)
Date: Sun Jun 04 2006 - 09:16:18 ART


No worry.

You have helped me quite a bit already.

I am familiar with the link:

http://www.cisco.com/en/US/tech/tk543/tk544/technologies_tech_note09186a00800
94612.shtml

the other is for ATM, so I am not sure if it is releavant.

I have done plenty of reading on Qos but not enough hands-on to see it at
work.

I will do that and see if everything falls better into place.

Thanks again.

Pierre
  ----- Original Message -----
  From: Petr Lapukhov
  To: Pierre-Alex
  Cc: ccielab@groupstudy.com
  Sent: Sunday, June 04, 2006 2:05 PM
  Subject: Re: QoS default-class - oversubscription

  Pierre,

  unfortunately, I have not much time to write thorough explanation,

  but I do recommend you to look at the following links:

  http://www.cisco.com/en/US/tech/tk39/tk824/technologies_tech_note09186a0080
0c9699.shtml
  http://www.cisco.com/en/US/tech/tk39/tk824/technologies_tech_note09186a0080
093d62.shtml#
  http://www.cisco.com/en/US/tech/tk543/tk544/technologies_tech_note09186a008
0094612.shtml

  They may give you some good insight on how CBWFQ actually works, and how
that
  "ureservable" bandwidth is used. Just remember the key idea - "bandiwidth"
is just
  a weighting factor for undelying WFQ algorithm.

  HTH
  Petr

  2006/6/4, Pierre-Alex <paguanel@hotmail.com>:
    I have labed point number 2 and I my statement is actually wrong!

    With the configuration as is, the router will refuse the commands because
the
    available bandwidth is actually 75% of the interface bandwidth .

    ( I think I got confused because when I did the test with shapping, I was
    abled to use 100% of the shapped rate. )

      ----- Original Message -----
      From: Pierre-Alex
      To: ccielab@groupstudy.com
      Sent: Sunday, June 04, 2006 11:38 AM
      Subject: QoS default-class - oversubscription

      I need a quick "yes" or "no" + quick optional comments on the following
    points .

      Theses are deductions I have made from several readings on CCO and
research
    in the group-study archives.

      I just want to make sure I am not inventing wrong rules here ...

      1. QoS bandwidth usage is calculated based on L3 packet size.

      2. On non-7500 + platforms: 25% of the interface bandwidth SHOULD be
left
    aside for overhead traffic, including Layer 2 (frame? overhead), control
    traffic (routing,cdp,snmp etc ...), and best-effort traffic. This traffic,
and
    in particular NON IP traffic, will be allocated to the DEFAULT-CLASS

      Because the 25% is a recommendation and is NOT enforced by the routers
when
    using absolute bandwidth values. Nothing prevents me from doing the
following
    and starve my overhead traffic.

      interface s 0/0
      bandwidth 100

      policy a
      class 1
         bandwidh 90
      class 2
           banwith 10
      clas-class default

      3. The command "maximum bandwith-percent xx" is only releavant when
using
    configurations with "bandwidth percent" .
      When left to the default maximum bandwith-percent 75 of interface
bandwidth
    and when using "bandwidth percent " type configurations, then the overhead
is
    taken care off.
      When changed to maximum bandwith-percent 100 then care must be taken to
    leave enough in the default class for overhead traffic.

    _______________________________________________________________________
    Subscription information may be found at:
    http://www.groupstudy.com/list/CCIELab.html



This archive was generated by hypermail 2.1.4 : Sat Jul 01 2006 - 07:57:31 ART