[ietf86-tech] Guestrooms & Public Wireless

[Warren Kumari]

On Feb 4, 2013, at 7:00 PM, Jim Martin <jim at daedelus.com> wrote:

> 	Guest room connectivity at the Caribe is both wireless (2.4G b/g) and wired in each room. The wired is real 100M ethernet, back to a switch on every other floor, that has a gig uplink to a central distribution switch. The wireless is 3 APs on every other floor, and is known to be spotty. 
> 
> 	The public space has 4-5 APs in the reception building (with the bar & restaurants) and one out at the pool. 
> 
> 	All of this ties into the same gateway box, that does billing/NAT/rate limiting (1M per user) all sharing a 20M pipe. 
> 
> 	We had the usual discussion with Lori, Keith and Angela about guest room connectivity and public space wireless. After a bit of discussion, they agreed that the easiest thing for everyone involved would be to have the IETF/IEEE take over all guest room and public space networks at layer 3, using the existing layer 2. VeriLAN will be on the hook for providing 24x7 support if users have issues, with the Caribe staff handling any physical issues.
> 
> 	From physical perspective, this is pretty straightforward. We'll be taking an ethernet handoff in their IT office (more on physical/cabling layouts shortly), with zero re-config on their side. We're literally moving the cable from the port on their gateway to a port on our switch. The emergency backout would be to move the cable back. 
> 
> 	The tentative plan is to switch over to our network at noon on the 9th and return to their network at noon on the 22nd. 
> 
> 	Now the gotcha of all of this is that every room is it's own VLAN.  Warren has been working on this, and believes we have a config that can handle all the rooms out of our usual pool, but it requires the BRAS/bridge-domains functionality that exists only in the MX series. Hence the reason for us to have a pair of MX80s rather than M7is this time around.
> 
> 	Warren, would you mind running us through your plan?

Sure -- the MX routers are a combination router and switch (yes, so are M series, but these are more so :-P)

So, we create a bunch of L2 (no L3) sub-interfaces, one per VLAN and set the encapsulation to type vlan-bridge. Config below:
interfaces {
    ge-1/0/0 {
        flexible-vlan-tagging;
        encapsulation flexible-ethernet-services;
        unit 100 {
            encapsulation vlan-bridge;
            vlan-id 100;
        }
        unit 200 {
            encapsulation vlan-bridge;
            vlan-id 200;
        }

This wil accept tagged frames on each VLAN and toss the VLAN tag. We then put all of these into a bridge-domain:
bridge-domains {
    BD_NONE {
        vlan-id none;
        interface ge-1/0/5.0;
        interface ge-1/0/0.100;
        interface ge-1/0/0.200;
        interface ge-1/0/0.300;
    }
}

The "magic" is that we set the vlan-id to none in the bridge-domain. 
The then add an interface that will have L3 configured on it to the bridge-domain (interface ge-1/0/5.0; and configure it thusly:
    ge-1/0/5 {
        description "To L3";
        encapsulation ethernet-bridge;
        unit 0;
    }

I have been assured that this works (Ron tested it in a lab), but I am also getting a loaner MX shipped to my house to try it so I can test it for myself….

W


> 
> 	- Jim
> 
> 
> 
> _______________________________________________
> ietf86-tech mailing list
> ietf86-tech at daedelus.com
> http://www.daedelus.com/mailman/listinfo/ietf86-tech

-- 
Eagles soar but a weasel will never get sucked into a jet engine