Why are there 13 root servers?
Thirteen root server identifiers are a historical result of early DNS packet-size constraints and naming design. Modern DNS no longer has the same practical limit, while the established A through M identities remain in use.
The big picture
A recursive resolver starts at the root when it needs help finding the next authoritative layer of DNS.
or network
identities
.com, .org
name server
Introduction
The DNS root zone is configured with 13 named authorities, from a.root-servers.net through m.root-servers.net. Those names are identifiers for root server services, not a count of physical machines.
Today each identity can be delivered from multiple network locations using anycast. A resolver can therefore reach one of many instances while still querying the same root server address.
The historical limit
The number 13 came from a concrete engineering constraint. Original DNS over UDP was restricted to 512-byte messages. As the root server set grew, a priming response containing the root name server set and address records had to remain within that envelope.
- By 1993 there were eight root servers.
- The root servers were renamed into the
root-servers.netzone so DNS name compression could save space. - Four more identifiers were added in 1997, bringing the total to 13.
- Adding another identifier at that time would have exceeded the 512-byte design budget for important responses.
Modern DNS reality
The old 512-byte UDP limit is no longer the defining constraint it once was. EDNS allows larger DNS messages, IPv6 has been added to the root service, and the operational system has evolved considerably.
The key distinction is between 13 root server identifiers and the much larger number of operational instances that serve those identities around the world.
Anycast and scale
Anycast lets the same root service IP address be announced from multiple locations. Internet routing then directs a query toward a reachable announcement according to routing policy and topology.
- The A through M identities remain stable.
- Each identity may have many distributed instances.
- Failures at one site do not imply failure of the entire root identity.
- The deployment can grow without inventing new root server letters.
Common misconceptions
There are only 13 physical servers.No. There are 13 named root server identities backed by a much larger distributed deployment.
DNS can never have more than 13 root server identities.No. The historic 13 came from the operational and packet-size constraints of the period, not a permanent protocol law.
One organization runs the entire root.No. The 13 identities are operated by 12 independent organizations.
Further reading
The RSSAC FAQ contains the historical explanation and the byte-level reasoning behind the 13-identifier design. IANA publishes the canonical list of the A through M root authorities and their service addresses. For resolver bootstrap behavior, use the current priming BCP.