How Internet Works

The key to how the Internet works lies in understanding what it is: The Internet is a network of networks. It is not an easily defined single object.

The Internet is the world's largest distributed system; it was designed and engineered for redundancy (it has an abundance of routes and connections) and resilience (it easily recovers from a mishap). The Internet is not a single company or a group of companies, nor even a single network. It is a worldwide mesh or matrix of hundreds of thousands of networks, owned and operated by hundreds of thousands of people in hundreds of countries, all interconnected by about 8,000 ISPs (Internet Service Providers). No single organization controls the Internet; not the U.N.; not the biggest ISPs; and the Internet has long since outgrown control by the U.S. government.

The Internet is different from other major services. Electricity tends to be provided by a single company in each geographical area. The ``last mile'' of telephone service to the customer is usually owned by a single company. But in general there is more than one Internet provider in any locale, and there are usually many paths from a local provider in one area to a provider in another area.

When you, the user, look at a web page through the Internet, many things happen along the way. There are various ways to get from your house or office through the ``last mile'' to the Internet: modem dialup, ISDN, DSL, cable modem, wireless, leased line, etc. These various technical methods may provide speeds anywhere from very slow (a few hundred bits per second) to very fast (billions of bits per second). All these access methods are onramps to the information superhighway.

In order to transmit text or pictures, your data is chopped up into small packets which are routed through the Internet. But first they have to go from you to your local ISP, or the equivalent piece of the Internet inside your organization (an intranet). This local ISP is a possible point of failure. If something goes wrong at your local ISP, it may look to you like the Internet is broken. It's not. Only one small piece of it is broken. The rest of the Internet, with its portals and stock portfolios and shops and reams of scientific data and plethora of information and people on it will not break because one ISP does.

To reach a web server, your local ISP sends your packets of data to another ISP, which may send them to another ISP, or through an Exchange Point or a National point access (NAP) or local point access (LAP) to get to another ISP. Thus your packets pass through a chain of ISPs through nodal points to reach their destination. Your packets may pass through fiber optic cables in the ground, satellites in the sky, undersea cables, or radio links. They may travel at speeds including T-1 (1.544 Mbps), T-3 (45Mbps), or faster (or slower). The internet Protocol (IP) ties all of those links together, enabling your packets travel through the Internet.

Eventually your packets arrive at the web server, and the web server sends responses back along a similar path (almost definitely not the same one). Any of these Internet providers can have problems (congestion, broken link, power outage, broken computer, etc.), which may cause the web server to seem slow or unresponsive to you. But the web server is broken only if the web server is actually broken. Problems in intervening parts of the Internet do not break the web server, which may well be accessible to other people, and may become accessible to you as soon as the various Internet providers route your traffic around problems.

Much rerouting in the Internet is dynamic, and happens automatically. (Imagine you are driving up the California coast and come to a sign that says that there has been a mudslide. You drive inland, north on another road, perhaps rejoining the coastal highway again. You have changed your route dynamically.) Some rerouting isn't automatic. In particular, the biggest ISPs, frequently called backbones, cover vast geographical areas and carry large proportions of the Internet's traffic. A failure in a backbone or in one of the major interconnection points between them can affect many Internet users. And such a problem may take some time to be resolved, as the biggest ISPs often prefer to manually examine changes in major routes before implementing them.

But longstanding observation of the Internet indicates that ``some time'' is normally at most few hours, even in the face of the biggest problems.

Internet providers use the same methods for routing packets for electronic mail or file transfers or remote login or voice or video. People tend to be quicker to notice slowness in accessing web pages, so we have used accessing a web server as an example.

There are other key pieces of the Internet, most notably the root name servers. Nameservers translate domain names, such as www.ripe.net, into the IP addresses, such as 193.0.0.195, that are used by the Internet protocols in carrying your packets through the Internet. The root nameservers handle the most basic part of that translation, which is finding nameservers for the top level domains (TLDs), such as NET, COM, ORG, EDU, GOV, FR (France), JP (Japan), AU (Australia), or PE (Peru). The root nameservers are widely spaced in both geography and in Internet topology, so that a failure in one cannot readily affect another. The root nameserver operators have also cooperated in extensively testing their software, hardware, and capacities, and they all know how to reach each other in case they perceive problems.

The rest of the Domain Name Service (DNS) is distributed among hundreds of thousands of nameservers for the various domains. For example, there are nameservers for ORG, and then there are nameservers for MIDS.ORG. Every domain is supposed to have at least two independent nameservers, and most do (another instance of redundancy). In any case, a failure in a single nameserver may make a particular domain temporarily inaccessible, but it will not affect the Internet at large.

The decentralization of the Internet is one of its biggest advantages and one of its most basic features, designed into its protocols from the beginning and tested in practice over many years. If one piece breaks, that doesn't mean the Internet is broken. And decentralization requires cooperation, so the various ISPs and IXes and the like are accustomed to cooperating with one another to fix and prevent problems. It is this decentralization and cooperation that has permitted the Internet to grow faster for longer than any other technological phenomenon in history. It is important for you, the user, to understand how decentralization makes the Internet work, so that you will know that the Internet is actually very hard to break.