1

In the link below, it is said that there are near one million prefixes for IPv4. https://bgp.he.net/report/netstats

I want to know that it is a must for ISP routers to cache all these million IPv4 entries? Do any differences exist between tier 1 or tier 2 ISPs when answering this question?

Actually, which Routers over the internet need to cache the whole IPv4 routing table? Where are these router? Who own them?

1
  • 1
    Remember that those routers are very large and expensive, with lots of resources (RAM, CPU, etc.). Even many businesses that have an AS and peer with multiple ISPs will sometimes have the full Internet routing table in their WAN routers.
    – Ron Maupin
    Commented Feb 3, 2020 at 15:06

2 Answers 2

2

I want to know that it is a must for ISP routers to cache all these million IPv4 entries?

No. However, ISP's & CDN's wanting to run within a DFZ (Default Free Zone) will usually hold all of these prefixes.

Do any differences exist between tier 1 or tier 2 ISPs when answering this question?

The only difference you may see here is the source of the routes. A lot of Tier 1 providers will not do much Multilateral & Bilateral public peering as its more cost effective to sell their customers IP Transit to reach their AS and or setup PNI's (Private Network Interconnects) with Tier 1/2 Providers & CDN's.

Actually, which Routers over the internet need to cache the whole IPv4 routing table? Where are these router? Who own them?

Like mentioned above, usually only large ISP's and CDN's that operate with the DFZ for traffic engineering purposes will hold these routes.

4
  • I have gone through the BGP world report for the USA and Germany. bgp.he.net/country/US bgp.he.net/country/DE The largest number of prefixes for an ASN in those countries is 155k and 84k respectively. I couldn't find any ASN which announces something like 400k at all. How's that? Why those expensive and million-routes routers are not reported on the website. Don't they run BGP?
    – A.A
    Commented Feb 4, 2020 at 6:51
  • 2
    @A.A What you're looking at is 'Prefixes Announced' which is purely how many prefixes an AS is transitting for its customers. HE (AS6939) runs one of the worlds largest International IP Transit Networks and as we'd expect, announce a lot of aggregated prefixes to the world. This page doesn't show the size of the AS's routing table but purely the statistics on how visibile that AS is to the world.
    – ditrapanij
    Commented Feb 4, 2020 at 7:00
  • So, what are the other prefixes which are placed in the routing table and make it a million-entries table if they are not the prefixes announced by BGP? I thought BGP has the biggest impact on routing table size!
    – A.A
    Commented Feb 4, 2020 at 8:00
  • 3
    @A.A It is, those prefixes you see under ‘Prefixes Announced’ are ONLY routes that are announced by AS6839 and their downstream ASes. There are hundreds of thousands of routes that don’t touch AS6939. For example, an Indian ISP could announce plenty of routes that have no affiliation with AS6939.
    – ditrapanij
    Commented Feb 4, 2020 at 8:22
2

For the purposes of this post, I will use the following definitions of the tiers.

  • A tier 1 ISP is one that relies purely on peering and does not use transit.
  • A tier 2 network (may or may not be an ISP) is one that uses a mixture of peering and transit.
  • A tier 3 network (may or may not be an ISP) is one that relies purely on transit.

(many definitions of "teir 1" also require that the peering be settlement free, but that is not relevant to the issue at hand here and is not possible to reliably determine from outside)

You can't go pointing a default route at a peer, at least not if you want them to remain a peer in good standing. So the core routers of a tier 1 ISP will need to be default-free. To reach the whole internet without a default route, a full internet routing table is needed. Therefore the routers in the core network of a tier 1 ISP will need to carry the whole internet routing table.

A tier 2 or tier 3 network on the other hand can choose if they wish to filter the routing tables from their upstreams and peers and use a default route pointing at one of their transit providers for the residual traffic. This saves them router resources, but has the downside that if their upstream provider gets in a peering dispute they may lose access to parts of the Internet.

Many ISPs will split their network into "core" and "access" portions. The "core" side will have a full internet routing table, while the access side will have default routes pointing at the core side.

1
  • Thank you, your answer is completely helpful, I wish I could select two correct answers.
    – A.A
    Commented Apr 5, 2020 at 11:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.