[Ietf-62] Re: Network Requirements and Equipment

[Jim Martin <[e-mail]>]

Dennis,
	Again, thanks for the advance work.... Comments inline below...


On Feb 6, 2005, at 12:49 AM, Dennis Fazio wrote:

> --On February 5, 2005 05:12:42 PM +0100 Jim Martin <[e-mail]> 
> wrote:
>
> [...]
>> streaming from each of the breakout rooms, we'll need an additional
>> ethernet In addition to what we need for the APs.
>
> There should be two ethernet drops in each of the rooms, one for WiFi, 
> one for wired unicast audio. We will be heading to the Hilton 
> Wednesday (9th) to verify wiring and jack locations in each room.

	Great. Could you also verify a few other things while you're there?
	- What do we have running down from the roof to the 3th floor cloak 
closet? I seem to remember two pair of Cat5, but I'm not positive. If 
it's fiber, what is it terminated with?
	- What do we have running up from the demarc room in the car park? I 
honestly don't remember if this is fiber or copper. Also, does it 
terminate on the 3rd floor closet? And, as always, if fiber, what does 
it terminate in?
	- I'm assuming the gig from Onvoy will be delivered in that demarc 
room in the car park. Do we know what type of connector they're 
delivering on? Last time we had to scramble for the cable. What gig 
standard are they going to be using? ZX-IR or LR?


>>
>> 	We have the bar and lobby area on the ground floor, 5 rooms (6 if you
>> count the NOC) on the 2nd floor, and the usual Grand Ballroom split 
>> into up
>> to 7 Salons with airwalls (ugh!), plus the perimeter rooms (Directors 
>> Row
>> 1-4, and Board Room 1-3, as well as Duluth and Rochester) on the 3rd 
>> floor.
>
> It wasn't clear from Marcia's chart that we needed access in 
> Director's Row 1-4 and the Board rooms. I don't think we included them 
> in our Access Point estimate. Will there be events in those rooms that 
> require coverage?

	I'll let Marcia answer that conclusively, but often these are offices 
and such that people would appreciate the wireless, but don't demand 
it.

>
>> We also have the Ramsey room, which I can't seem to find on the 
>> floorplans,
>> but I'd expect it to be on the 3rd floor (anyone know for sure?).
>
> Ramsey is the room marked "Retail" adjacent to the Carver room on the 
> second floor.


	Ah, thank you!

>
>> several Onvoy upstreams as well as UMN. ChrisL has arranged a pair of
>> Junipers from Lenny, and Dave has a "pile of Cat 3750's" (20-30) from 
>> UMN
>> that we can use.
>
> Dave was of the opinion that we could use the 3750s as access routers 
> and the Junipers would not be needed, so we should settle on what our 
> preferences are for that.

	Right, this is exactly the discussion I wanted to get to. I'm hesitant 
to use the 3750s as our sole source of routing for several reasons. The 
biggest one is probably their limited ability to handle encaps/decaps 
duty. This means that if we need to bring up a tunnel to anywhere, for 
any purpose, we're SOL. It also means that they'll be more limited as 
to being able to act as a PIM RP, since this also involves a bunch of 
encaps/decaps (think register packets). While we're not doing /tons/ of 
multicast, we might as well build something solid :-) Also, the v6 code 
for the 3750 was just released something like 10 days ago ... it seems 
to work, but I not sure I want to bet the IETF on it :-) Finally, we 
really like to keep full BGP tables, so we can do optimal egress 
routing (which makes even more sense with lots of upstream peers, like 
we have this time). I've not tried this on a 3750, but I'd be worried 
about memory and CPU utilization.

	So, that all said, I personally feel much better using the Junipers.

	How's this for a proposed design. We place one M10 in the roof room, 
one in the car park. The roof one gets the wireless external (if that 
works out) and the car park gets the external Gig. We then place a pair 
of stacked 3750s in the 3rd floor cloak room (which I'll call the core 
switches). We run two pair from each of the M10s to the stack, 
terminating one in each physical switch, using 802.3ad link 
aggregation.  That way, we can't go down due to any one single failure. 
If we want to be paranoid, we can dual attach any fanout switches back 
to both core switches, and do 802.3ad (or cisco EtherChannel) to ensure 
that even the edge stays up in a single failure.

	From a routing perspective, we could ether drag all the vlans out to 
the Junipers, and do the usual VRRP trick, or we could use the Junipers 
solely as external routers and have the 3750's do all our internal 
routing.  I'd kinda like to do the latter, but I'm open to other 
thoughts on the matter!

	What do you guys think?

	- Jim

_______________________________________________
Ietf-62 mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/ietf-62