RE: cut-through or store-and-forward

From: Andrew (arousch@xxxxxxxx)
Date: Sun Jul 23 2000 - 03:31:40 GMT-3


   
   cats use store and forward by default.
   At 09:17 PM 7/22/00 -0500, Price, Jamie wrote:
   
     I was under the belief that Cat's use cut-through until the error
     rates reach a certain level (I dont know what the predefined level
     is) and at that point they cut over to store-and-forward. They
     remain in store-and-forward until error rates drop below the
     cutover level at which point the catalyst switches (no pun
     intended) back to cut-through.
     I believe (I will look into this more and get back to everyone)
     that the error level that causes the switch from cut through to
     store-and-forward is configurable. Then again I could be dreaming
     :)
     As I said I'll check into it more and get back to you.
     Jamie
     -----Original Message-----
     From: mark salmon
     To: Ccielab
     Sent: 7/22/00 10:29 PM
     Subject: Re: cut-through or store-and-forward
     Excellent explanation but a bit misleading. The processor you are
     referring to is an ASIC not a central CPU and it is on the SUP
     card.
     THe 6xxx uses a crossbar matrix so it eliminates the problem you
     mentioned.
     "Jason T. Rohm" wrote:
>
> I can't speed for the others specifically, but the 5xxx series
     (and
> presumably the others) use a hybrid switching type.
>
> Explain:
>
> A pure cut through switch switches all packets w/o error checking
     (other
> than runt).
> Pure Store-and-forward, copies the entire frame to the switch
     processor,
> checks it out, then copies the frame to the outbound port(s).
>
> A catalyst 5xxx, will immediately copy the frame to ALL active
     ports
     and the
> switch processor. The switch processor will check the frame for
     errors
     and
> destination, then inform all the non-destination ports to drop
     the
     frame
> from their send queues.
>
> The result is a hybrid switching type that gives you all the
     benefits
     of a
> pure store and forward, but incurs less switching delay because
     the
     second
> frame copy is eliminated.
>
> The disadvantage of this method is that it requires a lot of
     back-plane
> bandwidth and large buffers on the individual ports.
>
> -Jason T. Rohm
> jtrohm@athenet.net
>
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On
     Behalf Of
> alfred zhang
> Sent: Saturday, July 22, 2000 3:25 AM
> To: ccielab@groupstudy.com
> Subject: cut-through or store-and-forward
>
> Hi
>
> Can anybody help me?Does a Catalyst switch(for example
> 6509,4003,2948G) use cut-through switching by default and
> store-and-forward based on errors?
>
> Thank you and best regards
> alfred
>



This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 08:23:57 GMT-3