Re: WFQ

From: John (jgarrison1@austin.rr.com)
Date: Sun Jul 13 2008 - 11:43:32 ART


Narbik,

Wouldn't you consider the way "WFQ" treats traffic based on virtual time a
weighting mechanisim?

Thank you for the formula. I think I am starting to get it. The formula you
gave me is how WFQ weights the packets, and does not determine the bandwidth?
I don't get the set number, is that a constant that is used in all WFQ ip
precedence algorithms? Badnwidth is determined by "bandwidth /
X(0+1)+X(1+1)........X(7+1) " depending on how many packets are in the queue
and what their precedence is?

Thanks to all for the info
  ----- Original Message -----
  From: Narbik Kocharians
  To: Anthony Sequeira
  Cc: Joseph Brunner ; John ; ccielab@groupstudy.com
  Sent: Saturday, July 12, 2008 9:05 PM
  Subject: Re: WFQ

  WFQ without the weight should be fair-queuing; the weight aspect of the WFQ
comes in when traffic is set with IP Precedence levels.
  The formula that WFQ uses is as follows:

  If packets are 1500 Bytes and the IP Precedence is set to 0, then:

  [32384 / (IP Precedence + 1)] X 1500 = 32384 / (0 + 1) = 48,576,000

  The 32384 is a set number.

  If packets are 1500 Bytes and the IP Precedence is set to 1, then:

  [32384 / (IP Precedence + 1)] X 1500 = 32384 / (1 + 1) = 24,288,000

  Note the packets with IP Precedence of 1 appear to be half the size of
packets with IP Precedence of 0 and therefore they will receive twice as much
bandwidth as the packets that are set with IP Precedence level of 0.

  On Sat, Jul 12, 2008 at 4:01 PM, Anthony Sequeira
<Anthony_Sequeira@skillsoft.com> wrote:

    In the context of the CCIE Lab, not much to worry about here. WFQ is one
    of those "legacy" queuing mechanisms that does a good enough job to be
    the default for "low speed" interfaces still. It will indeed prevent a
    "top talker" from squeezing out a "low talker". <I think that was a
    Seinfeld episode!>

    There are some configurable parameters like the Congestive Discard
    Threshold and the number of Dynamic Queues, but one of the main problems
    with this queuing mechanism is that it does not support fixed bandwidth
    guarantees. If you want to ensure your voice traffic gets 256K across
    your WAN, this tool falls short.

    Class Based Weighted Fair Queuing came along and addresses the
    shortcomings of this mechanism. And then Low Latency Queuing further
    improved upon that. These are the Lab topics I would focus on.

    Interestingly, another place you see WFQ used today is in a LLQ config
    for the "catch all other traffic" class of class-default.

    I hope these comments helped you.

    Anthony J Sequeira
    #15626
    www.freeiestuff.com

    -----Original Message-----
    From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of

    Joseph Brunner
    Sent: Saturday, July 12, 2008 4:38 PM
    To: 'John'; ccielab@groupstudy.com
    Subject: RE: WFQ

    Much to learn about google have you still...

    http://nislab.bu.edu/sc546/sc441Spring2003/wfq/wfq.htm

    Much Mystery there will be until you search with keywords correct

    Answers to questions find you will on this link and others.

    google

    cisco wfq scheduler

    -----Original Message-----
    From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
    John
    Sent: Saturday, July 12, 2008 4:35 PM
    To: ccielab@groupstudy.com
    Subject: WFQ

    I don't know why I have a hard time getting this, but lets see if I have
    this
    right.

    WFQ will give low volume traffic "better treatment" then high volume
    traffic.
    It does this through some mysterious algorithm that I can't seem to
    find.
    WFQ
    is the default for all interfaces under E1 speeds. WFQ works with ip
    precedence and RSVP. WFQ is flow based. I cannot change the behaviour
    of
    WFQ
    through configuration. There are some QOS features I cannot configure
    on an
    interface with WFQ.

    Thats what I get. I would appreciate comments on anything I missed or
    got
    wrong

    _______________________________________________________________________
    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

    No virus found in this incoming message.
    Checked by AVG - http://www.avg.com
    Version: 8.0.138 / Virus Database: 270.4.9/1548 - Release Date:
    7/12/2008 7:40 AM

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

  --
  Narbik Kocharians
  CCSI#30832, CCIE# 12410 (R&S, SP, Security)
  www.Net-Workbooks.com
  Sr. Technical Instructor



This archive was generated by hypermail 2.1.4 : Mon Aug 04 2008 - 06:11:54 ART