Flat OSPF network, or single area OSPF networks are real. In fact most of the OSPF network today deployed, is flat OSPF networks. But how many routers can be placed safely in an OSPF area ? Any number from the real world OSPF deployment ? I will share in this post.
Let me explain what it is first and then will share you some numbers from the real network which I engaged recently.
As you might know, OSPF has two levels of hierarchy. Backbone and Non-Backbone areas.
Why Non-Backbone Areas are used in OSPF?
The reason is scalability and manageability. At least in theory. I don’t see so many multi area OSPF design though I teach in very detail in my CCDE classes. But that is for the exam purpose.
There are some very large scale networks use OSPF for scalability, so, IP but satellite (Sometimes called an Access POP) POPs are in Non-Backbone area they place.
But there is manageability aspects of having multi area OSPF design. They group their slow speed access and metro or aggregation networks in different OSPF areas and place high speed backbone/core routers in a backbone OSPF area (Area 0).
But, we generally forget to talk about the complexity of the Multi area OSPF networks. When people have Multiple areas, at the area boundary (ABR), they usually summarize the IP prefixes. I am not saying this is bad. But managing the summaries and filters increases the overall configuration of the network.
How many routers in real deployments flat OSPF network can accommodate?
You might see in textbooks , don’t place more than 40 – 50 routers in a backbone area. When the number of routers increases , place them in a separate areas.
This is not wrong, but this was old. 20 years before when you heard this, you could understand the main reason as hardware of the routers. They couldn’t handle the topology change of the network, they couldn’t handle the SPF CPU usage, they couldn’t keep all the prefixes in the memory and so on.
But of course, hardware is more advanced as of 2017. Routers can handle , in fact they handle , much more routers in a flat OSPF design.
Recently I engaged the project of an Access ISP, which has 700 POPs , but only 100 IP POPs, no MPLS (thus, BGP everywhere) , and around 200 Routers in the OSPF domain.
You could expect to see the affect of the topology change in 200 nodes routers in OSPF. But wait, these guys had ‘ Prefix suppression ‘ everywhere on the network. So they only had loopback interfaces on the routers in the routing table. They don’t have any problem with OSPF. Memory , CPU , routing table size , manageability , complexity everything is fine.
What next in this particular network ?
We will enable on this network MPLS, on top of that Traffic Engineering in the second phase. They won’t have any problem with Traffic Engineering because they have full topology visibility, thanks to flat OSPF design. They will do Traffic Engineering in a distributed fashion by using CSPF on the routers, but unless we have specific problems such as dead lock, global path optimization etc. it is fine.
So, 200 routers , with prefix suppression in place , in flat OSPF network design is working totally fine. Also read my ‘ Study Tip ‘ post to understand, why these numbers will not be important after maybe 5 – 10 years.