[c-nsp] Guaranteed bandwidth THROUGH our network

[Rick Kunkel]

Hello,

Hypothetical situation here, which will become real in fairly short order:

Let's say that I've got a customer paying premium prices for guaranteed
bandwidth.  He's connected to a 100 Mbps Ethernet port on a switch of ours
(along with several other customers, on different ports), which is trunked
via 802.1q to a 7206VXR NPE-G1 GigabitEthernet port.  Also connected to
that 7206 is a Gigabit Ethernet connection to a "Tier 2"  Internet Service
Provider, through which the customer will be receiving service.

This particular setup isn't necessarily what we would do in an ideal
situation, but that's what we're working with.

Cheesy ASCII art (apologies to variable-width viewers):

C1, C2, C3, etc. = Customer1, Customer2, Customer3, etc...

     (INTERNET)
         |
         | (1 Gbps Ethernet)
         |
   |-----------|
   |   7206    |
   |-----------|
         |
         | (1 Gbps 802.1q trunk)
         |
|-----------------|
|      Switch     |
|-----------------|
  |  |  |  |    |
  |  |  |  |    |
 C1 C2 C3 Cn... | (100 Mbps Ethernet)
                |
       "PREMIUM" CUSTOMER

The premium customer wants a guaranteed 50 Mbps to the Internet, at least
as far as we have control over.  So, if possible, I need to make sure that
if there is contention for bandwidth across these 2 traversed Gigabit
links, his is a minimum of 50 Mbps.

What is the best way and/or technology to address this?  Specifics are
cool, but I'm as willing as the next guy (hopefully the next guy is cool)
to RTFM.  More than anything, I'm looking for a direction here.  I'm
familiar with rate-limiting and the rudiments of ATM traffic
classification, so this won't be all new.  But I'm just not sure in what
area to begin my reading here.

Direction is much appreciated.

Thanks,

Rick Kunkel