[c-nsp] daisy-chain hop limit

[Keegan Holley]

On Mon, Mar 14, 2011 at 1:42 PM, Paul Wozney <paul at wozney.ca> wrote:

> On Mon, Mar 14, 2011 at 03:31, Nick Hilliard <nick at foobar.org> wrote:
>
> > On 14/03/2011 09:46, Marco Regini wrote:
> >
> >> i have to daisy-chain 28 cisco switch, my humble is only about
> >> spanning-tree (rapid-pvst).
> >>
> >
> > Here's what happened one hospital which did something like what you're
> > planning to do:
> > http://www.cio.com.au/article/65115/all_systems_down/
>
>
> Scalability?  none.
> Performance?  poor.
> Stability?  questionable.
>
> Additional cabling isn't terribly expensive, and it is a capital expense at
> that so once it is in you can recoup the benefits for years to come.  Get a
> core switch to aggregate your 28 access switches - if your cabling
> infrastructure is copper this is pretty easy, and if your runs are greater
> than 100m then I'm sure someone on the list could recommend a good fiber
> aggregation switch.
>

My favorite is the Juniper EX4200-24F.

>
> The story Nick linked is a good one to illustrate the potential pain of
> exceeding the recommended STP diameter, and you should also read the SRND
> on
> campus network design.  I know it looks like it is written for large
> enterprise organizations, but the same approach works in the SMB designs.
>
> http://www.cisco.com/en/US/solutions/ns340/ns414/ns742/ns815/landing_cOverall_design.html
>
>
Another idea is to create a ring of a few core switches and then create the
standard switch blocks connecting to two of the switches in the ring.  This
kind of design could work well with a rapid type spanning tree (MST or
RPVST) and careful planning if there are limitations of running the cabling
to one single place such as in a campus environment.  It's not safe to have
that many switches daisy chained though.  I'm just curious how the OP ended
up at this design.