The International network of networks known collectively as the Internet consists of private, public, and research computer networks connected by major corporate and goverment service providers. This patchwork of services together forms an unruly, but unified, whole which has structure on many levels. The World Wide Web forms a structure, albeit a changing one, by logically linking together widely disparate sites according to content-based critereon. UseNet, the internet news network, arranges data from sites according to a heirarchichal subject listing. Gopher sites link sites by content in a fashion similar to the Web. These are logical arrangements, and are independent of the underlying physical arrangement of the net.
The Internet is often portrayed in the newsmedia as a vast landscape of major corporations and shady characters, university professors and mom-and-pop bulletin boards connected by telephone lines and sattelites. The vastness of it is implied; there is rarely a statistic given. The most common summary seems to be the nature of its growth rather than any metric of size. As with any complex software system, it is most easily viewed heirarchically. There are a relatively small number of central service providers who carry the traffic from place to place. There are regional hubs which transfer this international traffic to other large networks. These large networks tie together gateway computers of service providers. Corporations and private individuals then connect to these service providers. These are the groups who ultimately receive and transmit information. Because it is often too expensive or slow to send every messsage through the main switching points, there are a myriad of iterconnections at the smaller provider and service level, and some individuals and corporate nets allow traffic from one service to "tunnel" through to other services. This level of interconnction does not today carry the bulk of the traffic, as the large service providers are taking an increasing share of the rapidly growing traffic. Hence, one may view the internet topology as a multi-layer network of nodes, each layer representing a certain level of service and each node representing a gateway to another network. Since we only mention the gateway nodes, we are truly looking at the "internetwork" topology and not the global network topology, which would include all of the source nodes as well. This is a slightly distorted picture of total network connectivity, since some providers may have millions of source nodes ( i.e. compuserve) I have yet to find anything that even purports to be a map of a significant portion of the internet. It has probably been a long time since any such thing actually existed. But we can still cobble together a sort of map by examining the maps of the major networks involved and understanding the main points of interconnection. For that is the true nature of the internet: it is the interconnection between vast numbers of different networks, and the top hundred or so nets carry by far the bulk of the traffic. Is knowledge of the configuration at any one time of any value? I have always found it to be so. For this interconnection is so often the topic of uninformed conversation and debate, and often this debate is taking place at very high levels with very little clear view of how the system has come into being and is currently configured. And to the user of the system, a little understanding of its structure can go a long way towards making its use more pleasant and successfull. As with any story, it's best to start at the beginning.
To have an internet, you must first have a network. Computer networking began in a proprietary fashion: digital equipment machines ran DECNet, IBM mainframes ran SNA, and so forth. Each of these companies, and many of their clients ran huge, international networks based on these protocols in the 1970s. The concept of open networking came out in about 1970 with the introduction of the OSI referenece model. ARPA adopted NCP, the OSI-based predecesor to TCP, in 1972 on the early Arpanet connecting UCLA, UC Santa Barbara, SRI, and University of Utah. The network was developed by Bolt, Barenek, & Newman (BBN). Over the period between 1972 and 1985, when the Arpanet project wound down, the internet grew from a few hosts to several thousand. By 1985, many of the sub-networks were larger than the whole net had been ten years before. Berkeyely's implementation of TCP/IP set the standard for open protocols, and Berkeley's network became one of the main hubs of activity. Other major connection points arose near Stanford, and in New York at Columbia and Cornell. Vint Cerf describes this portion of the Internet's history in a SF Chronicle interview.
The military had been using sattelite links and dedicated telephone lines for some time to transmit orders to missile bases and to manage communications. With more and more military money going into research projects through the ARPA, it became apperent that widely distributed groups had need for tools to allow them to communicate findings and share documents and datasets over a wide area. The Arpanet was born, and one of the first maps of the early internet would be this one of Arpanet circa 1974. Nothing here yet Following the breakup of Arpanet in 1985, many service providers began springing up. Most of them were formed from the people who originally built the internet, or had worked for Bell Corps at some point. These companies now form the core operating systems of the internet.
UuNet technologies was founded in 1987. Its backbone map may be seen at www.uu.net/bbone.html
Performance Systems International (PSI) began by building part of what is now the major infrastructure in the northeastern US. Their current map is at www.psi.net/psi-tech/topology.html
Carrying data across the pacific is done by cable and sattelite. Sprint provides major service in the Pacific.
The New Zealand internet is run by a consortium of research organizations. On the North Island, their main feed is carried by Sprint into Waikato, where it serves the TUIA research network, Midland, the ICONZ, and KCBBS. Midland feeds the service providers Cybernet and the Internet Group, while TUIA feeds INCA and IProLink. KCBBS feeds Web World.
Main APCCIRN link into Melbourne
NORDUnet is an international network operator that provides services and international connectivity to the Nordic National Networks in Denmark, Finland, Iceland, Norway and Sweden. Nordunet has their map at www.nordu.net/operational-data/lines.gif . They operate for the following Networks:
Russia is a major growth site for the internet. Sprint provides extensive service there, as shown by the map at www.rosprint.ru/about/s_networks/imx/map1small.jpg .
In preparing this site, I used two primary methods and alot of random poking about. The first method was to search Yahoo and Lycos for "Internet Service" and the various contries and organizations. The second method involves the use of the Unix traceroute program. I picked a host in the country I was interested in learning about. I then ran a traceroute from my host in California ( well.com ) to the target host, and noted the major carriers on the path. I then found their Web sites and searched them for maps. This was by far the most instructive excercise. I do not encourage people to use this technique in general because it does generate quite a bit of traffic, relying as it does on RIP timeouts along the route. I was also aided by several excellent journal articles, including the ACM article on connectivity in North Africa and the Digital Video Magazine article on ISDN providers in the US, which contained a fine list of major US carriers. The whole project was triggered by requests from my wifes family for international access, since they are the most widely scattered family I have ever seen and they all want to exchange e-mail. The Godfrey family consists of:
