[c-nsp] 4900M vs. 4503 for core

[scott owens]

>
> Message: 3
> Date: Wed, 27 Jan 2010 13:42:43 -0500
> From: "Jason Gurtz" <jasongurtz at npumail.com>
> To: <cisco-nsp at puck.nether.net>
> Subject: [c-nsp] 4900M vs. 4503 for core
> Message-ID:
>        <A92EAF652EC423438D55C14C60771C8701F3E5AD at exchgsrv.nputilities.local
> >
> Content-Type: text/plain;       charset="us-ascii"
>
> We are doing a long overdue redesign of our network as part of a voip
> implementation, hopefully ending up with a collapsed core w/routed access
> layer.  A consultant has proposed the 4507 as access switches and a pair
> of 3750-E switches as the core.  The 3750-E seems a strange choice to me
> for a few reasons and I'm thinking a pair of 4900M or 4503 switches would
> be a better fit looking forward.
>
> We are a smaller shop (7 access switches including the datacenter) with
> 100Mb desktops and a mix of 100/1000 for servers.  Switch-to-switch trunks
> are 1Gb.  The number of access switches is very unlikely to change and we
> could, in the future move to a 10Gb.  The 4900M solution would save a
> non-trivial amount over 4503 with Sup6.
>
> Is there anything glaringly wrong with choosing the 4900M using twin-gig
> based connections to the access layer over the 4503 Sup6 and 46xx line
> cards in our situation?
>
> ~Jason
>
>
> What the 3750Gs - I would pick the "G" over the E/10GB  until you need 10
GB ... the price will have dropped by then -
is an clustered switch stack that you can run redundant etherchannels to ;
one link to one 3750, one link to the other.

Yes, yes, yes ... if you have to reboot the 3750 stack you lose all
connectivity as both ( or more ) members reboot.
It switches, it routes , it port channels.
jumbo frames, teaming, ....

Save your money until you need [ unless you are going to run 10GB to 10GB
links where a single client to server connection will exceed 1 GB you can
just as easily port channel a bunch of 1 GB links and get almost the same
effect.