[nsp] hardware

[Shaun R]

+Jon
This is really my first attempt at a network that is not just flat.  I
though about just doing the subnet on each port but i'm not really sure how
i would route that down from the core switch to the switch in that rack.  My
provider said they would assign a /29 so that there router and my
switch/core would talk and than i would take my ips and route them.  If i
wanted to subnet at each customers port would i then need to setup a /29
between my core and edge switch?

Doing subnets per customer port would make managing alot easier, at least in
my mind it would.  I suppose it's not really that much diffrent.


+Stephen
If i grouped customers into a single vlan than wouldnt things like broadcast
and arp traffic go to all of the customers with in that vlan.  I wont say
the name of the company but i worked for a company that has a completely
flat network, and have over 200 servers and a /19.  The problems with
customers affecting other customers was terrible and the amount of
broadcast/arp traffic that gets totalled up for there bandwidth usage is
alot.  They have customers that run bw monitors and show that they only push
10GB's that month when the companys showed 20GB.  That company has choosen
to have a poor network and it costs them customers.  I dont want to have
that problem, i want one customer to affect another customer as least as
possible.

~Shaun


----- Original Message -----
From: "Jon Lewis" <jlewis at lewis.org>
To: "Shaun R" <mailinglists at unix-scripts.com>
Cc: <cisco-nsp at puck.nether.net>
Sent: Monday, July 05, 2004 4:58 PM
Subject: Re: [nsp] hardware


> On Mon, 5 Jul 2004, Shaun R wrote:
>
> > Thanks for your reply.  My main concern in this design is to make sure
that
> > I have room for growth.  I don't want to implement something that's
going to
> > give me problems when I double/triple my customer base.  That's why I
was
>
> We just did this (just?..around a year ago) and chose to use 3550-48's as
> the aggregation switches (deployed in or around the customer racks).  At
> the moment, we're also using 3550-48's as the core swithes.  Right now,
> the agg switches have 100mb back to the core switches which then have FEC
> into the routers.  At some point in the future, the core 3550's will be
> replaced with gigE switches, the agg switches will get gigE uplinks to the
> core, and the FEC from the core switches to the routers will go gigE.
>
> > It sounds like the switches I have chosen are a good choice.  I am still
> > concerned about how I am doing to do the subnet/vlan'ing.  It seams like
if
> > I had say 1000 customers, and a sub interface for each customer/subnet
that
> > it would be a huge pain in the ass to manage.  I mean the config would
be
> > huge.
>
> Why do a vlan for every customer when you can run their switchport in
> layer3 mode?  We only do vlans when necessary.  i.e. MPLS VPN vrfs,
> customer insists on having multiple systems in the same subnet.  The
> latter one I suspect could have been done with BVI interfaces on the
> 3550s, but I didn't think to test that pre-deployment, and I knew vlans
> would work...and we had to do some of them anyway.
>
> ----------------------------------------------------------------------
>  Jon Lewis                   |  I route
>  Senior Network Engineer     |  therefore you are
>  Atlantic Net                |
> _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
>