DNS and the structure of the internet

Overview

Recall questions on the internet and the world wide web, the Domain Name System and its hierarchy, resolving a domain name, DNS caching and cache poisoning, authoritative name servers, and the role of ISPs and the backbone.

What you need to know

Answer each one before you reveal it.

Questions from this subtopic

Each one with the model answer underneath.

1. 1

   What is the internet, and how is it different from the world wide web?

   Model answer

   The internet is the global infrastructure: a network of independently owned networks, joined by routers and running common protocols, over which data can travel between any two connected machines. The world wide web is one service that runs on top of it - a collection of interlinked pages and resources, identified by URLs, requested using HTTP and HTTPS and viewed in a browser. Email, file transfer, voice calls and online gaming also run over the internet but are not part of the web. The internet is the roads; the web is one kind of traffic on them.

2. 2

   What problem does the Domain Name System solve?

   Model answer

   Routing works on IP addresses, which are numeric and change when a site moves to different hosting. People cannot remember numeric addresses and would have to be told a new one every time a site moved servers. DNS solves this by mapping memorable, human-readable domain names onto the IP address currently serving them. Users and links refer to the name, so the owner can change hosting, move to a different provider or use several servers, and only the DNS record has to be updated - every existing link keeps working.

3. 3

   Domain names are read in an unusual direction. Explain the hierarchy in a name such as shop.example.co.uk.

   Model answer

   The hierarchy runs right to left, from the most general to the most specific. At the far right, above everything, is the unnamed root. Then uk is the top-level domain. Then co is the second-level domain within uk. Then example is the domain registered by the organisation. Then shop is a subdomain that the organisation itself created and controls. Each level is responsible for delegating and knowing about the level immediately below it, which is what makes the system a hierarchy rather than one enormous list.

4. 4

   Describe step by step how a domain name is resolved into an IP address.

   Model answer

   The browser first checks its own cache, then the operating system's cache, for a recent answer. If there is none, the request goes to the resolver operated by the internet service provider or the local network. The resolver checks its cache. If it still has no answer, it asks a root name server, which does not know the address but replies with the address of the name server responsible for the top-level domain, for example uk. The resolver asks that top-level domain server, which replies with the address of the name server responsible for the registered domain. The resolver asks that authoritative name server, which holds the actual record and returns the IP address. The resolver caches the answer and returns it to the client, which can now open a connection to that IP address. If any server in the chain already has the answer cached, it returns it immediately and the remaining steps are skipped.

5. 5

   Why are DNS answers cached, and what stops a cached answer being wrong forever?

   Model answer

   They are cached because the full chain of queries takes time and would have to be repeated for every single request, adding delay for the user and an enormous load on the root and top-level servers. Caching at the browser, the operating system and the resolver means the overwhelming majority of lookups are answered locally in a fraction of a millisecond. Every record carries a time to live, a number of seconds for which the answer may be kept. Once that expires the cached copy is discarded and the next request triggers a fresh lookup, so a changed record spreads across the internet within the time to live. A short time to live means changes take effect quickly but caching helps less; a long one means better performance but a slower changeover when a site moves.

6. 6

   Why is DNS organised as a distributed hierarchy rather than one central lookup table?

   Model answer

   A single table would be a single point of failure: if it went down or was attacked, no name on the internet could be resolved. It would also have to handle every lookup made by every device on the planet, which no single machine could do, and it would be a bottleneck adding delay for users far away from it. A single organisation would have to authorise every name change worldwide. Distributing the data means each organisation is authoritative for its own names and can change them itself, load is spread across many servers, failure of one server affects only part of the namespace, and answers can be served from a nearby machine.

Other ways to work this

The same content, four other ways.

What is inside Levo

Six ways to work the same material.