Layer three to the closets

[thessick]

We are considering moving to layer three links to the closets using EIGRP stub on 3750's. Does anyone have experience with this? Comments? Recommendations?

Thanks

Tom

[vkapoor5]

The 12.2(25)SEE IP Base images for both Cat3560 and Cat3750 support EIGRP Stub Routing. All IOS versions before this require IP Services or Advance IP Services feature sets to enable EIGRP Stub Routing.

EIGRP stub provides full routing support for only those Layer 3 devices directly connected to the switch and not for routers downstream of the switch. Available in the IP Base images for the Catalyst 3750 and 3560 switches. Complete EIGRP routing support is available on those switches in the IP Services and Advanced IP Services images.

[leighharrison]

Hi there Tom,

I've designed and built around half a dozen routed access solutions recently.

As well as double throughput, there are also great advantages like easier troubleshooting, easier traffic management, easier QoS, etc., etc.

I try to design and build this way wherever I can now - in my oppinion, it is a much superior design than switched.

Out of interest, why are you choosing EIGRP? On my implementation where the customer has bought standard software, I have used RIPv2 and it works a treat.

The only thing to be aware of is that the standard software image on a 3750 does not support multicasting.

Hope that helps,

LH

Please rate all posts

[thessick]

We have used EIGRP since the network was implemented. Have you experienced the layer 3 convergence issues noted in the other post?

Initially we are planning a 6509 with four 3750 stacks at a remote site but would like to do the same locally with about 55 closets.

Recommendations on the stub configuration switches?

Thanks

Tom

[mchoo2005]

Be wary of how much you're planning to grow, too. If you're catering for a fairly large user population that will require stacking multiple 3750s, make sure that you test the Layer-3 convergence within the stack when the (stack) Master switch fails. Based on our earlier tests, the Master switch reelection and Layer-3 convergence could take minutes, not seconds.

[thessick]

Thanks for the info. We are planning about 55 closets with dual connected 3750 stacks. It appears to be a better option than spanning tree. What is the affect on the overall network? Also, have you seen a middle 3750 fail leaving two separate stack with the same network?

Thanks

Tom

[Aaron Harrison]

Hi Tom

You shouldn't see a split stack with 3750s as you connect the top switch to the bottom one as well to complete a resilient loop.

I've personally not seen any convergence take the minutes mentioned above - provided you sensibly optimize timers and reduce the size of the tables things converge very fast.

Regards

Aaron

Please rate helpful posts...

[leighharrison]

Hi there,

I agree with Aaron. The convergence times are pretty quick on these boxes and mean time between failure is something silly like 30 years.

There are 2 ways that you can use the stackwise cables on the back of the switches. You can either do one after the other from the top to the bottom and then have a long (1m, 3m, etc) cable returning from bottom to top. Or you could connect the short cables in such a way that you don't need the ong cables. Have a go at working it out - have a look at this link (figure 3-7):-

http://www.cisco.com/en/US/products/hw/switches/ps5023/products_installation_guide_chapter09186a00801b1742.html#1151483

Good luck with it and let us know how it works out for you.

LH

[glen.grant]

The other option is to use spanning tree but use rapid spanning tree and route at the 3750's which conververges within 2-3 seconds that we have seen and seems to work pretty well now . The other poster is correct if the 3750 switchstacks are wired correctly you should never have a split stack .