Using IPv4 anycast for load balancing

IPv4 anycast is actually a very old technique, and it works as follows. One set of address space is announced in multiple physical locations using BGP. All data sent to this IP address block will travel to the “nearest” location BGP hop-wise, because the router will do BGP balancing and choose the path that has the shortest BGP ASN hops in it.

This works pretty well for stateless UDP-related services – for example, DNS UDP-based load balancing and distribution. This is what first-class providers such as Neustar Ultra Services (former UltraDNS), CriticalDNS and many root TLD operators use to offer first-class DNS services.

Many ISPs will block any route announcements that are smaller than /24.


Comments

  1. Josh says:

    You just announce 192.168.1.0/24 prefix in USA and Europe. Make sure your upstreams are not filtering it and announcing further. Then you are set.

  2. Josh says:

    Sorry, almost forgot, check out Quagga (better than Zebra) for BGP open source speaking software.

  3. Ariel says:

    Don’t you need a /24 PI space to do proper IPV4 Anycast ?

  4. Kevin says:

    Only TLD and ccTLD can get PI space for anycast anyway. I think with IPv6 this may change and become more common.

  5. TaranBaran says:

    Do you need your own ASN or it’s perfectly fine announcing this block without own ASN?

  6. Jeff says:

    I think you can announce this prefix without your own ASN…. it should be no problem

Leave a Reply