As far as OSPF, I know that Cisco recommends that you should not have an OSPF area with more than 90-100 routers. Is there such restriction or recommendation for ISIS areas or levels?
ISIS Routing – Maximum Number of ISIS Routers in an Area or Level
is-isrouting
Related Solutions
OSPF Backbone
Why is area 0 the backbone area in OSPF? Why must all other areas connect to it?
This is explained very well in RFC 3509, Section 1.21:
1.2 Motivation
In OSPF domains the area topology is restricted so that there must be a backbone area (area 0) and all other areas must have either physical or virtual connections to the backbone. The reason for this star-like topology is that OSPF inter-area routing uses the distance-vector approach and a strict area hierarchy permits avoidance of the "counting to infinity" problem. OSPF prevents inter-area routing loops by implementing a split-horizon mechanism, allowing ABRs to inject into the backbone only Summary-LSAs derived from the intra-area routes, and limiting ABRs' SPF calculation to consider only Summary-LSAs in the backbone area's link-state database.
OSPF is usually considered a link-state protocol. What some people miss is that OSPF uses both link-state protocol and distance-vector protocol algorithms.
- Routes within the backbone, or a non-backbone area are computed as a link-state protocol does (ref Dijkstra's algorithm).
- When OSPF must carry non-backbone routes through the backbone, it uses some distance-vector behavior (i.e. parts of the Bellman Ford algorithm) to propagate Type3 LSA metrics into non-backbone areas.
Simple example of OSPF's distance-vector behavior:
<-- Area 5 --><-- Area 0 --><-- Area 4 -->
R5-----------R1-----------R2------------R3---------------------R4
Cost 3 Cost 5 Cost 7 Cost 12
LSA--> LSA-->
Type3 LSA Type3 LSA
{From R1} {From R2}
R5 cost is 3 R5 cost is 8
Consider what happens to a /32 Loopback route for R5.
- R5 sends a Type1 LSA containing the /32 Loopback
- R1 (Area 5 ABR), is connected to Area 0; it translates the Type1 LSA into a Type3 LSA with a cost of 3.
- R2 (Area 4 ABR) receives R1's Type3 LSA (metric 3) and changes the metric to R5's Loopback, based on R2's cost to R1. Now R2's Type3 LSA for R5 has a cost of 8. This is the distance-vector behavior I mentioned above.
Requiring all non-backbone routes to go through the backbone is a loop-prevention mechanism.
Connecting non-backbone OSPF areas at an ABR
If 2 areas aren't connected through area 0 (discontiguous), how does OSPF behaving as a link state protocol increase the possibility of routing loops?
As we saw above, OSPF uses distance-vector behavior to send routes through the Area 0 backbone. Distance-vector protocols have well-known limits, such as the count-to-infinity problem. OSPF would be vulnerable to the same issues, if we didn't have boundaries on its behavior.
1RFC 3509 describes Cisco IOS's ABR behavior
You can have multiple ABRs which connect an area to Area 0. With a stub area, a router in the area could choose the best ABR in the path to the destination, but with a totally stubby area, it will just choose the nearest ABR.
For instance, if you have two ABRs for your area (Area 1), each ABR will have, at least Area 0 and Area 1 interfaces, but one of them could also have an Area 2 interface, or it could be directly connected to an Area 2 ABR with a low cost, but the other ABR would have a higher cost to get to Area 2. It would make sense, when sending traffic to Area 2, to send it to the ABR with the lowest cost to Area 2, but a totally stubby area doesn't know how to do that.
Best Answer
In theory, just like for OSPF, there is no limit. There will be a practical limit, based on the routing protocol traffic generated, bandwidth, and router resources.
Some people have networks that work with 1000 routers for either OSPF or IS-IS, but that would be rare, and it is probably a network that needs to be redesigned. I actually worked on one, and it took over an hour to converge. Eventually, it was broken up into ASes, using BGP between the ASes, and OSPF withing the ASes.
In practical terms, you will probably see no more than a couple of dozen routers in any OSPF or IS-IS area. More than that would probably be symptomatic of a network that needs to be redesigned.
All the areas, except the backbone are on the same level, so you can have an almost unlimited number of areas at the same level, but they will be separate areas.