Inhouse MPLS/VPLS

[MichaelM55]

New building, new switches, new design?

imagine a huge building with different areas (simple office rooms, laboratories, library, canteen, lounging areas, lecture halls, ...), so basically a kind of campus.

And now take 17 10508/10512 switches to interconnect all of this.

Till now we have plenty of rather small buildings throughout the city, easily interconnected with OSPF and VRF, simple design, each building has its own purpose; physically separated from the rest.

In the future, everything will be hrown into one pot. But my users basically want to have that old setup in the new building, i.e. their own vlans throughout the whole building.

But tagging each one through the whole core?  No way!


Therefore I played with VPLS the last weeks. What do you think about the following setup?

- OSPF interconnecting all switches, setup all VLANs and subnets as usual, nothing special here
- But for getting that subnet beyond the switch; so instead of tagging, take VPLS to connect all rooms. (So MPLS/VPLS for "tagging", only.)

I´m still facing a problem. Imagine that VPLS tunnel. There I have to actually put in a cable to get the traffic through it, so on a 400 port switch, I have to setup the corresponding  port with VPLS?
But wait, couldn´t I simply take one port as "uplink", setup the VPLS config on this and take the rest of the ports and tell them to use this (a kind of gateway port)? (Hopefully it´s clear what I want to do)

Perhaps, someone did something like this before and give some hints. I don´t know whether that´s a bad idea at all...

Or do something completely different?

[paulgear]

MPLS/VPLS seems like the wrong tool for the job to me.  It would complicate the solution without adding any extra functionality.

Have a look at GVRP - it provides a way to trunk VLANs through all the appropriate core ports without ever having to configure them individually.  Turn it off for access ports and on for uplink ports on your edge switches, and turn it on for all the interconnects and downlinks on your core/distribution.  Then sit back, watch the packets flow, and enjoy.

On one campus deployment i did (about 1200 end nodes across 30 or so VLANs), we didn't even need OSPF, because all routing was done by one IRF stack in the core.