I'd like to seek out everyone's opinion on the use of IS-IS in a service provider network.
Cisco highly recommends the use of IS-IS in SP networks. Their OSPF book tells SP users to consider IS-IS. Their ISP Essentials books basically says that EIGRP is used most often by beginners since it is so easy to set up. It goes on to say that OSPF is better because it forces good IGP design which helps scaling. Finally it says that IS-IS is used experienced users because it allows for even better scaling and more configuration options than either of the other two. From my experience I would have to agree with that. I have less experience with IS-IS but I've been studying hard on it for about a month both in and out of my lab.
The network in question currently runs OSPF in a poorly conceived layout. Area 0 spans 3 POPs, one joined with fiber and the other joined via radio links. They are separated by about 80 miles. Area 0 has been pushed out to all devices including devices in the access layer. This means that a AS5300 in one POP has full routes to all the Lo0 IPs in all the other POPs, all the PtP subnet everywhere, etc. Random other areas are in use across the network as well. The IGP isn't enormous by any means but it is about 95% crap. The network will be undergoing serious core upgrades in the next few months at the 2 main POPs and will eventually be able to be called a professional network.
We have met with about half a dozen SEs to discuss our design and configuration and they all strongly recommended IS-IS for our network. The faster convergence time will be greatly relied upon by the VoIP traffic that cross our network not only for phone over cable and ADSL2+ users but also for customers with standard land lines (our software switches rely on this network for the switch to switch communication between POPs).
Few people have even mentioned EIGRP for an application such as this. Most people agree on OSPF and IS-IS as the contenders. I agree with Cisco and vote for IS-IS.
Does anyone have any applicable knowledge or experience that they would like to share? One of my questions involves the use of areas with a single POP. Should a POP be a single area or should we break up the IGP even more and place each large access device or types of access devices in a dedicated area? I'm thinking about placing the CMTSs in an area, the ATM routers for DSL in an area, the dialin in an area, etc (or each large access device in its own area. Every access device will be dual-homed to large core routers. Should I think more at where summarization would/could occur or the individual number of routes a device might stick in an IGP? We always summarize routes on access servers and ATM routers for this reason. I've mapped out the physical network and know where the L1 routers are (access devices) and L1-L2 devices are (all edge and core routers in our design). The design contains no L2-only routers.
Any input would be appreciated. I'm confident I could start rebuilding the IGP with IS-IS today, but I would like to hear everyone's thoughts on IS-IS from an operational standpoint.
Thanks J