The past several years have seen significant efforts to keep local Internet communications local in places far from the well-connected core of the Internet. Although considerable work remains to be done, Internet traffic now stays local in many places where it once would have traveled to other continents, lowering costs while improving performance and reliability. Data sent directly between users in those areas no longer leaves the region. Applications and services have become more localized as well, not only lowering costs but keeping those services available at times when the region's connectivity to the outside world has been disrupted. I discussed the need for localization in a previous paper, "Internet Mini-Cores: Local connectivity in the Internet's spur regions." [1] What follows here is a more specific look at a particular application, the Domain Name System (DNS).
Most Internet applications depend on the DNS, which maps human-readable domain names to the Internet Protocol (IP) addresses computers understand. Two Internet hosts may have connectivity to each other but be unable to communicate because no DNS server can be reached. This article examines the placement of DNS servers for root and top-level domains and the implications of that placement on the reliability of the services these servers provide in different parts of the world. It is not a "how-to" guide to the construction of DNS infrastructure and does not contain recommendations on DNS policy; it is rather a look at the placement of DNS infrastructure as currently constructed.