I'm seeking out advice on connecting multihomed access & distribution devices back to a redundant core. Currently I'm working with a mix of methods.
1) A /28 or /27 VLAN was created on each core router to contain the aggregate links to the lower layered devices. The access/distribution devices have a L3 interface in this subnet.2) A /30 is used between the core routers and the access/distribution layer devices.
3) A /29 with HSRP is used across the core routers on L3 VLAN interfaces and across the Ethernet interfaces on the access/distribution devices. Each core router has VLAN with a L3 address in the /29 and a common standby IP also in the /29. The access/distribution layer device has 2 Ethernet links back to the core that are 1Q trunks. That VLAN has a L3 address in the /29. This only works on a few devices, namely switches used for aggregation or routers with L2 ports like ENET HWICs.I'm trying to figure out what the best approach is in this service provider environment. I personally favor #2. I believe this fits in better with a clean hierarchical IGP. I'm moving from OSPF to IS-IS in the very near future and would like to better utilize areas than they have been in the past (OSPF area 0 is touching every single device). #1 would make it difficult to separate IS-IS areas I think, especially if I use the VLANs for inter-connection VLANs for more than one type of device that should be in separate areas.
IPs aren't a concern. We're migrating from a public to a private IP infrastructure for the heart of this service provider. This will make ARIN happy.
On a related topic, I'm also soliciting advice on the basic configuration moving forward for edge and aggregation devices. Like I said earlier, we're moving to a pure IS-IS environment with a redundant core. All devices will be multihomed to the redundant core routers. All devices whether it be an access server, router terminating ATM PVCs for DSL, or a pair of aggregation switches that mulithome CMTSs will be running IS-IS and will be advertising the local routes back to the core (some with summarization like the access servers and ATM routers). The access/distribution layer devices are set up in one of 2 ways right now. Either their default route is originated in the IGP or they use a static default route (and in some cases still receive an default route from the IGP but don't use it). Should I originate a default route and remove the static default, should I use the static default route, or should I have a higher cost static route as a backup (rather useless I think)? Is there a best practice on this one that I can cite?
I just thought of another quick question. Do many people use (or find useful) 1Q trunks in a service provider network? I only have 2 1Q trunks in this entire network. Personally I would rather use L3 routing between capable devices wherever possible. I suppose if I use a pair of switches for aggregation (EMI code on 3750s or 4948s for example) I could use a pair of 1Q trunks between then for HSRP purposes. Do any other uses come to mind?
Thanks J