Metro Ethernet with Dual HQ Routers

[Philip Denton]

Network gurus, I've got a client migrating from OPT-E-MAN over to a Metro Ethernet solution.  They have 20 remote sites, all of which are Layer-2 adjacent to their 2 headquarters routers (each in a separate DC).  Remote sites are configured as EIGRP stubs.

The proposed Metro-E design is to encapsulate each remote site into a (carrier) VLAN and then hand it off at one of the two data centers on a trunk link.  After reading  through "Ethernet Access for Next Generation Metro and Wide Area Networks" it appears that the recommended architecture here would be to use subinterfaces to create a bunch of point-to-point routed links between each HQ router and each remote router, mimicking a Frame Relay environment.

Unfortunately, extending two VLANs to each remote site would double the cost of those circuits so that's not an option but the carrier has agreed to extend all remote site (VLANs) to both HQ routers so I'm curious what the next best solution would be.  I figure I could configure a slew of /29s and use EIGRP off-sets on the subinterfaces to do some primitive load-balancing.  Maybe HSRP would be better?

Any ideas would be much appreciated.  Please let me know if I left out any critical info.

[Philip Denton]

I just touched base with my client's carrier and for a small charge we can re-provision the WAN as either any-to-any OR the dual-HQ point-to-multipoint.

What are the pros and cons of each?  Seems to me that any-to-any with HSRP on the default gateways would be the easiest but I can't help but feel like I'm missing something here.

TIA,

Phil

[Giuseppe Larosa]

Hello Phil,

a dedicated Vlan for each remote site with three devices over it speaking EIGRP is a better solution from a routing point of view.

It gives you also better control with the use of offset-lists on HQ routers

A single Vlan with any to any connectivty looks like great, but if there are issues on the VPLS it may be difficult to troubleshoot and to fix.

Of course, you need also to take in account user traffic flow paths: if there is little communication between remote sites and most of traffic of each remote site is exchanged with the two central sites I would go with the separate Vlan for each remote office approach.

Indeed two different Vlans for each remote site would give you even better control of routing.

Hope to help

Giuseppe

[Philip Denton]

Giuseppe, thanks for the speedy reply.  Unfortunately two VLANs per site is not an option, since our carrier would charge us double the access fees per site.

Would you then recommend running "plain" EIGRP and offset lists on the HQ routers rather than running HSRP instances on each VLAN?

Once again, I appreciate the speedy reply but if anyone else would like to weigh in please do!

Thanks and have a great weekend,

Phil

[Giuseppe Larosa]

Hello Phil,

I see so a single Vlan for each remote site, three routers on it the two HQ routers and the remote site router.

>> Would you then recommend running "plain" EIGRP and offset lists on the HQ routers rather than running HSRP instances on each VLAN?

HSRP is a First Hop Redundancy Protocol not a routing protocol,betweeen routers managed by the same company, same working group, the use of a real routing protocol has to be preferred.

So yes I would not waste time configuring HSRP groups on HQ routers I would just use offset-lists to build remote sites to HQ direction load sharing.

Actually the HQ routers, using different offset-lists on interfaces pointing to core, could also achieve load sharing in the opposite direction from HQ to remote sites.

Hope to help

Giuseppe