On DNS architecture: One of the DNS series

DNS is currently one of the most indispensable servers on the Internet, and every day we surf the web with the help of DNS. The DNS server can resolve the domain name for us, locate the e-mail server, and locate the domain controller in the domain ... In the face of such an important server role, we need to do some in-depth research, this article tries to explore the DNS architecture, so that people can better understand the principle of DNS.

The main task of DNS is domain name resolution, that is, to translate the computer name into an IP address, so that we can directly with the computer name easily associative memory for network communication without the memory of those dull and obscure IP address. Now let's give you a question of how computer name resolution is done on the Internet before DNS appears. This issue is clearly meaningful, describing DNS RFC882 and 883 appeared in 1984, but the November 1969 Internet was born, before the advent of DNS, the Internet pioneers are communicating with each other with IP address? Of course not, but the early internet was really small, with only 4 hosts on the Internet, at the University of Utah, Stanford, UCLA, and Santa Barbara, California, which had only hundreds of hosts on the internet throughout the 70 's. In this way, solve the problem of name resolution can be used a very simple method, each host using a Hosts file can be all the Internet to resolve the host. This hosts file is now in use, the path is in the \windows\system32\drivers\etc directory, as shown in the following figure is a Hosts file example, We can see very clearly in the diagram that hosts file resolves www.baidu.com to 202.108.22.5.

On a small scale internet, the use of Hosts file is a very simple solution, in general, the host administrator of Stanford University updates weekly hosts files, the other host administrators regularly download the updated Hosts file every week. But obviously this solution in the rapid expansion of the Internet is not very applicable, even if the internet now has 100 million hosts, think about it, if everyone's computer must have a host of 100 million Hosts file! Oh, is not going to collapse!

Internet managers found a successor to the hosts file in time-dns,dns's design requires a distributed structure that allows the host to decentralize the management of data while the data can be used by the entire network. Decentralized management is conducive to easing the bottleneck of a single host, easing the flow of pressure, but also make the data update become simple. DNS is also designed to use a hierarchical namespace to name hosts to ensure the uniqueness of host domain names.

The DNS design requirements you have seen, I will explain the following specific. DNS predecessor Hosts file is a complete decentralized resolution scheme, each host is responsible for name resolution, this method has been denied by us. So can we use a fully focused parsing solution? That is, the world only a Hosts file, Internet users use this file for name resolution! This plan I listen to still have the merit of place, at least everybody is freed up, do not need each computer to update that hosts file, the world as long as the only hosts file maintenance good Bob. Actually carefully consider, there are many problems, such as the host file hosts will become a performance bottleneck, facing huge flow pressure, and the results of each domain name resolution to be updated through this file, updated speed can not be imagined too timely. Therefore, DNS also does not adopt this completely centralized parsing scheme.

At present, DNS adopts a distributed parsing scheme. Specifically, the Internet Management Committee stipulates that the domain name space is the root of the resolution of the server all, that is, the roots of the Internet all the domain names have a complete right to resolve! Wait a minute, there are readers to ask questions, that this root server is not equivalent to the world's only hosts file it? Oh, do not worry, the root server with a simple operation, the change of this structure. What operation does the root server use? Delegate! The following figure is a diagram of the root server delegation, as shown in the following illustration, where the root server delegates the domain name resolution rights of COM endings to other DNS servers, and all future domain name root servers that end with COM are not responsible for parsing, and the server being delegated is responsible for parsing. And the root server is also the end of the net,org,edu,gov, such as the domain names are delegated, these delegated domain name is called the top-level domain, each top-level domain has a preset purpose, such as COM domain name for commercial companies, EDU domain name for educational institutions, gov domain name for government agencies, etc., This top-level domain name is also called the top-level organization domain name. The root server also carries out domain name delegation for different countries, for example, all the domain names that end with CN are delegated to China Internet Management Center, the domain name of JP End is delegated to Japanese Internet Management Center, CN,JP these top-level domain name is called the top-level geographical domain name.