From: Omer Ansari (omer@xxxxxxxxxx)
Date: Thu Jul 25 2002 - 20:33:26 GMT-3
Here's another nice example I ran into:
Curb Congestion with Custom Queuing
http://www.cisco.com/warp/public/784/packet/jan99/17.html
On Wed, 24 Jul 2002, Chris Hugo wrote:
> Example from Vegesna BOOK
> You have three protocols A,B,C.
> Your queuing needs are of the following distribution:
> A-20%
> B-60%
> C-20%
>
> Facts on the average packet size of packet:
> A-1086 bytes
> B-291 bytes
> C-831 bytes
> To determine byte-count perform the following algorithm
> 1. FOR Each queue divide the % of bandwidth you want allocated to the queue b
y packet size
> 0.01842 020619 0.02407
> 2. Normalize the numbers by dividing by the lowest number
> 1 11.2 1.3
> THE Result is the ratio of the number of packets that must be sent.
> 3. Fraction in any ration values means an additional packet is sent
> 4 In the example, the number of packets sent is one 1086-byte packet, 12 291-
byte packets and 2 831-byte packets.
> He goes into checking the whole thing but I am in a rush .sorrry.
> chris hugo
> Tom Larus <tlarus@cox.net> wrote: The 12.1 IOS docs on this subject note th
at getting the byte-count right is
> not as important now as it used to be, because now the custom-queuing system
> remembers the deficit when a queue is depleted early or when the last packet
> from a queue does not exactly match the configured byte-count, and accounts
> for this the next time the queue is serviced.
>
> Does this mean we really don't need to worry so much about the packets sizes
> used by different protocols, or that if we don't get it right, our result
> will be less wrong than with 12.0, but would still be wrong? It would nice
> if we could be accurate using the default byte-counts now.
>
> Does anyone have a guide to the approximate packet sizes of various
> protocol's packets? The default is 1500. I read that DLSW needs FST
> traffic to be no more than 512 bytes, and I see that a lot of people set lf
> to 1500 for TCP DLSW traffic. I would imagine that this is so traffic
> originating on a 4472 MTU token ring interface doesn't have to be fragmented
> at a 1500 MTU ethernet interface. I seem to remember seeing IPX set to 1024
> in a practice lab solution, but I don't know if it is right. Telnet tends
> to use tiny packets, but does anyone know what a good reference byte-count
> would be?
>
> Thanks,
> Tom Larus
> ----- Original Message -----
> From: "Tommy C"
> To:
> ;
> Sent: Tuesday, July 23, 2002 9:22 PM
> Subject: Re: FW: Custom Queueing bandwidth allocation
>
>
> > I think you'll also need the average packet sizes for each queue to come
> up
> > with the byte counts.
> >
> >
> > >From: "Ivan"
>
> > >Reply-To: "Ivan"
>
> > >To:
> > >Subject: FW: Custom Queueing bandwidth allocation
> > >Date: Wed, 24 Jul 2002 08:52:49 +0800
> > >
> > >Hi,
> > >
> > >Had a question on Custom queueing.
> > >When a question ask to use custom queueing and allocate different
> bandwidth
> > >to each queue, I"m not sure how to work out the bandwidth allocation base
> > >on the percentage.
> > >
> > >Example:
> > >
> > >allocate
> > >queue 1 to ip traffic with 50% bandwidth
> > >queue 2 to ipx traffic with 20% bandwidth
> > >queue 3 to sna traffic with 30% bandwidth
> > >
> > >
> > >With this, how should I specify in the
> > >queue-list 1 protocol
> byte-count
> > >command.
> > >
> > >Appreciate for any advice.
> > >Thanks.
> > >Ivan Lim
> > >
> > >
> > >
> > >------------------------------------------------------------------------
> > >Take part in the Penning on Squares Chinese Essay Writing Contest.
> > >Submit your masterpieces or vote for your favourite works, and win
> > >attractive prizes.
> > >Be there or be SQUARE!
> > >http://sea.litcontest.lycosasia.com
> > >
> > >cOntact @ Lycos
> > >= 10MB for email and filestore + lots of other goodies...
This archive was generated by hypermail 2.1.4 : Sat Sep 07 2002 - 19:36:44 GMT-3