Link state vs Distance vector for large networks

doctorlexus

I could have sworn that I heard Jeremy Cioara say, in CBT Nuggets, that link state routing protocols were not good for larger networks because the routing/topology table becomes incredibly large. But a lot of what I'm reading online lately is saying link state is preferable for large networks, whereas distance vector is better for smaller networks. If that's the case, why do ISP's use BGP, which is distance vector? I feel like I'm getting conflicting information.

In my mind, Jeremy's explanation made sense, since distance vector protocols only needed to know next hop information. And for that, it seemed like they'd be better suited for large networks because one didn't need to worry about overly large routing/topology tables. So either I misheard him, or I'm misunderstanding something fundamental here.

clarson

distance vector updates the whole routing table on each update. That is every 30 sec. with ripv2.
link state only updates the routing table when there is a change and then with only the changes. Lots less network traffic for a large network. But, then the router needs to recompute the routing topology.

networker050184

Speaking of ISP networks you're usually going to run a link state routing protocol and BGP. You can carry hundreds of thousands of prefixes in BGP both the internet routing table and internal customer prefixes. It's very stable and scalable, but it's also mostly slow to react. You then run your underlying IGP, OSPF for instance, to carry the next hops (usually loopbacks) for recursion. That allows your underlying next hops to converge quickly while not having hundreds of thousands of routes churn in BGP.

doctorlexus

Sounds more complicated than I anticipated. Looking forward to the CCNP ROUTE exam, as I should hopefully understand all of this better by then.