Link Fragmentation Calculation ?

Hi every1,

      If any one can help me on this subject . When we do LFI . To get the Fragment_Size or Delay we use the following calculation

Fragment_SIZE =  Bandwidth * Delay .       

My question is ,  the Bandwidth that we take in the calculation is it the 

1. configured Bandwidth on the interface   (As Configured by Bandwidth command on the interface)  ?

2. the Clock Rate . ?

3. the Actual interface rate. ?

4.  And In case of frame relay the CIR. ?

 

Thanks

Regards

Mazin

 

Comments

  • You use whatever the interface thinks its bandwidth is.  Just type "show interface <interface> | in BW" and use that.

  • <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">






    Go by actual rates.  Clock rate is actual
    transmission, "bandwidth" is an artificial marker.  Unless modified, they
    should be set the same!

     

    Don't forget mincir to accomodate anything bad.  Watch
    your details.

     

    Typically, it's 80 bytes per 64K of link
    speed.

     

    Also typically (real life) we don't fragment on link speeds
    greater than 768K.  That's because the fragmentation delay is longer than
    the transmission delay is.

     

    In the lab, do what you're told! 
    :)

     

    HTH,

     




     



    Scott Morris, CCIE4 #4713, JNCIE-M #153, JNCIS-ER, CISSP, et al.
    CCSI/JNCI-M/JNCI-ER
    Senior CCIE Instructor

    <?xml:namespace
    prefix = st1 ns = "urn:schemas-microsoft-com:office:smarttags" />[email protected]

     

    <?xml:namespace prefix = o ns =
    "urn:schemas-microsoft-com:office:office" />
    Internetwork Expert, Inc.
    http://www.InternetworkExpert.com
    Toll
    Free: 877-224-8987
    Outside US: 775-826-4344
    Online Community:
    http://tinyurl.com/6dmnsu
    CCIE Blog:
    http://tinyurl.com/2nxxaq


    Knowledge is power.
    Power corrupts.

    Study hard and be Eeeeviiiil......


     





    From: [email protected] [mailto:[email protected]]
    On Behalf Of Mazin
    Sent: Wednesday, December 24, 2008 7:17
    AM
    To: [email protected]
    Subject: [CCIE
    R&S] Link Fragmentation Calculation ?


    Hi every1,

          If any one can help me on this subject . When
    we do LFI . To get the Fragment_Size or Delay we use the following
    calculation

    Fragment_SIZE =  Bandwidth * Delay
    .       

    My question is ,  the Bandwidth that we take in the calculation is it
    the 

    1. configured Bandwidth on the interface   (As Configured by
    Bandwidth command on the interface)  ?

    2. the Clock Rate . ?

    3. the Actual interface rate. ?

    4.  And In case of frame relay the CIR. ?

     

    Thanks

    Regards

    Mazin

     




    Internetwork
    Expert - The Industry Leader in CCIE
    Preparation
    http://www.internetworkexpert.com

    Subscription information
    may be found
    at:
    http://www.ieoc.com/forums/ForumSubscriptions.aspx
  • My other question regarding Fragmentation.

    1. When we do fragmentaion on Multilink PPP.  The fragmentation calculation is done  based on the configured Bandwith of the multilink ppp interface or from somewhere else ?

    2. My second question regarding FRF.12 fragmentation. Lets consider that we have a hub and spoke topology e.g

    R1 is the Hub and R2 & R3 are the spokes.  R1 actual speed of the link connecting with the SP switch is 786Kbps . The service provider have provisioned us with 512K on DLCIs b/w R1 & R2 .

    Whereas b/w R1 & R3 we have been provisioned 128Kbps. Now when i will do fragmetentation calculation.

    Should I take the actual rate of the interface or the provisioned rate for the respective DLCIs?

    And do i have to take into consideration of some other parameters like CIR,Bc etc

    Thanks [:)]

    Regards

    Mazin













  • 1.  You fragment whenever there's a concern about overall
    latency affecting more critical traffic.  It's done based on whatever your
    bandwidth happens to be.  Multilink lines make things a bit more complicated.

     

    With links in general, it's 80 bytes per 64K of line speed.  We
    don't fragment things (real-life) on lines above 768K in speed.  That's because
    the fragmentation delay will be longer than the serialization delay of the
    line.  (e.g. you're making things worse by fragmenting).  In a lab, you do
    whatever you're told!  :)

     

    With multilink lines, let's say I have two 512K circuits bonded
    together.  A 1M line, I wouldn't fragment, but a 512K line I would.  Each link
    as it transmits individually is only 512K.  But the load-balancing of MLPPP may
    help us avoid the problem.  In other words, you look at what kinds of things
    you are sending, how much traffic there is, and determine whether you HAVE a
    problem before you try to fix it!  :)

     

    2.  Lowest speed is what I'd go with.  You don't want to set
    yourself up for a failure someplace.  Fragmentation values are unidirectional. 
    Fragmentation must be ALLOWED on both sides though.

     

    HTH,

     

     



    Scott Morris, 
    CCIE4 #4713, JNCIE-M #153, JNCIS-ER, CISSP, et al.

    CCSI/JNCI-M/JNCI-ER

    Senior CCIE Instructor



    [email protected]

     

     

    Internetwork Expert, Inc.

    http://www.InternetworkExpert.com

    Toll Free: 877-224-8987

    Outside US: 775-826-4344

    Online Community:
    http://tinyurl.com/6dmnsu

    CCIE Blog:
    http://tinyurl.com/2nxxaq



    Knowledge is power.

    Power corrupts.

    Study hard and be Eeeeviiiil......

     

     

     

    From: [email protected]
    [mailto:[email protected]] On Behalf Of Mazin

    Sent: Thursday, January 08, 2009 4:04 AM

    To: [email protected]

    Subject: Re: [CCIE R&S] RE: Link Fragmentation Calculation ?

     

    My other question regarding Fragmentation.

    1. When we do fragmentaion on Multilink PPP.  The fragmentation
    calculation is done  based on the configured Bandwith of the multilink ppp
    interface or from somewhere else ?

    2. My second question regarding FRF.12 fragmentation. Lets consider that we
    have a hub and spoke topology e.g

    R1 is the Hub and R2 & R3 are the spokes.  R1 actual speed of the link
    connecting with the SP switch is 786Kbps . The service provider have
    provisioned us with 512K on DLCIs b/w R1 & R2 .

    Whereas b/w R1 & R3 we have been provisioned 128Kbps. Now when i will do
    fragmetentation calculation.

    Should I take the actual rate of the interface or the provisioned rate for
    the respective DLCIs?

    And do i have to take into consideration of some other parameters like
    CIR,Bc etc

    Thanks Smile

    Regards

    Mazin







    Internetwork Expert - The Industry Leader in CCIE Preparation

    http://www.internetworkexpert.com



    Subscription information may be found at:

    http://www.ieoc.com/forums/ForumSubscriptions.aspx





  • Hi.

    I'm  proving LFI for generate packets with 15ms of delay in a multilink ppp link of 128kbps. My problem is that I do not  find information about this command "debug ppp multilink fragment delay". I want to know what mean "encsize" and the number "4"

    I'm confuse because I made two proves with the same configuration and  the results are totally differents

    Result 1:

    Apr 30 13:29:04.855: Se0/0/1 MLP: I ppp IP (0021) size 44 encsize 4
    Apr 30 13:29:04.875: Se0/0/1 MLP: O ppp IP (0021) size 85 encsize 4 via oqueue
    Apr 30 13:29:04.955: Se0/0/1 MLP: I ppp IP (0021) size 44 encsize 4
    Apr 30 13:29:04.995: Se0/0/1 MLP: I ppp IP (0021) size 44 encsize 4
    Apr 30 13:29:04.995: Se0/0/1 MLP: I ppp IP (0021) size 46 encsize 4
    Apr 30 13:29:05.007: Se0/0/1 MLP: I ppp IP (0021) size 85 encsize 4
    Apr 30 13:29:05.099: Se0/0/1 MLP: O ppp IP (0021) size 87 encsize 4
    Apr 30 13:29:05.107: Se0/0/1 MLP: O ppp IP (0021) size 1470 encsize 4

    Result 2:

    May 10 12:18:52.795: Se0/0/1 MLP: I frag C0003BA6 size 90 encsize 4
    May 10 12:18:53.035: Se0/0/1 MLP: O ppp IP (0021) size 84 encsize 4
    May 10 12:18:59.595: Se0/0/1 MLP: O ppp FULL-HEADER (0061) size 141 encsize 4
    May 10 12:18:59.631: Se0/0/1 MLP: I frag C0003BA7 size 128 encsize 4
    May 10 12:19:02.795: Se0/0/1 MLP: I frag C0003BA8 size 90 encsize 4
    May 10 12:19:03.035: Se0/0/1 MLP: O ppp IP (0021) size 84 encsize 4
    May 10 12:19:12.791: Se0/0/1 MLP: I frag C0003BA9 size 90 encsize 4
    May 10 12:19:13.035: Se0/0/1 MLP: O ppp IP (0021) size 84 encsize 4
    May 10 12:19:14.383: Se0/0/1 MLP: I frag C0003BAA size 50 encsize 4

     

  • Hello pat_rilp, the encsize may be the MLP header : 4 bytes long.

  • Thanks. Now that you answer to me, I can see, I don't realized, plop :).
    But what is the difference between : " I ppp IP (0021)" and "I frag" and why frag include the sequence number and the other not?.


  • I guess it is Input/Output. are interleaving packetson boths sides of the link?


Sign In or Register to comment.