Bandwidth Buzzers and Bells: Deciphering Network 'Technise' - International Networks

by Steve Lieke

Data flow is the key to a good network. Like a Network Operations Center (NOC), bandwidth availability not capability determines the speed of data flow. Theoretically any Web host service can offer global service. Yes, an archaeologist in South Africa should be able to download your web pages and order parts for a 1979 Toyota Land Rover through a satellite uplink, if one was available. Of course it might take 20 or 30 minutes for each page to download and cost more that the truck cost in the first place, but the capability is there.

It is commonly believed that the Internet came into existence as part of the United State's government's nuclear war fighting strategy. Originally conceived as a robust and fast communications network called the ARPAnet (Advanced Research Project Association Network), it was designed to help scientists and technology researchers communicate, in fact, it was created in direct response to the Soviet Unions launch of Sputnik, the first man made satellite. It was part of a whole series of initiatives taken by the US government to enhance science and technology development. Later, the multi-routed and redundant telecommunications lines, switches, and computers were discovered by the military, to be an ideal network to prevent Command Control 'de-capitation,' in the event of a Soviet 'first strike' against the continental US.

More realistically, a mechanic in Cape Town phones the local parts wholesaler, who in turn, searches the web and comes up with your company's 1979 Toyota Land Rover part supply web page. The majority of this communication takes place over standard telephone cables. The tendrils of the shared (peered) networks extend through telephone cables, network routers, underground sea lines and satellite communication links throughout the globe and are the system that links our parts wholesaler to your site.

The Internet routers, which guide data transmission, are similar to the routers that interconnect the 'server farms' inside a Network Operations Center -- in the topographic language of the Web, these routers are also known as 'nodes'. By broadcasting information about line availability on the net, they constantly let each other know if they are available to relay data. Today, most routers use a common protocol developed by Cisco called Border Gateway Protocol 4 (BGP4) to communicate. This protocol helps routers determine the shortest route between two points in the network. They determine this in terms of least number of nodes between the beginning point and end point of a data exchange - not in terms of distance.

The problem with these public networks is that they service a huge amount of traffic and the quality and speed of transmission varies greatly. Most large network operators offer a combination of redundant data delivery systems, both on private lines and through 'peering' agreements with large telecommunications companies. To help increase bandwidth capability, and thus availability , many network operators in the US offer 'one hop' coast-to-coast dedicated fiber optic connections between their Network Operations Centers, but no one can offer it worldwide -- yet.

The Internet was designed as a distributed network . Before its creation, telecommunications connections were linear; like Christmas lights, if one in the chain went down, the whole string went out. Instead of relying on one network connection to transmit data, the Internet works like a spider web, with several possible paths leading from one point to the other.

In this model, data is transmitted from computer to computer in a series of 'hops', going down the chain of computers and being redirected at each point to another computer. These network computers are known as routers. The routers in the system decide where to send the data based on an address (called and IP address) that travels with transmitted data. This way, as opposed to a linear network, even if one network connection is closed, data can find another route through another series of links. It may take more time but, in the end, it is a surer method of transmission than a single direct line.

By paring down the number of nodes traveled, a dedicated 'one hop' network connection may help speeds transmission: a NOC in Silicon Valley might connect directly with another data center or relay point in New York - in this case even though data has traversed over a thousand miles it has never really (or virtually) left the network operators network! A company offering a private 'one hop' data transmission service is really saying, 'we offer an express line for our data that goes direct to major Network Access Points (NAP's) without making local stops.' Bandwidth is a commodity in high demand, and like deep sea telephone cables in the past, various large private interests are now extending these 'one hop' cables worldwide. However, that is the future. Until these lines are in place, data transmission must still compete with the vast amount traffic transported by public telephone services. Even here there are ways to enhance service.

To help our archaeologist in Cape Town, a large company based in the US could set up a 'data caching' server in England. This server would store much of the basic data that is required for Web page download, such as static web site design features. This data would then be close at hand for access from Europe and a little closer to our South African archaeologist. The 'data caching' server only needs to contact the home server in the US when it needs specific updates or unusual data.

While our archaeologist waits impatiently, sweating in the noon sun, the vital data she requires is making the leap across the continental US, then the Atlantic, through Africa and then to the local parts wholesaler in Cape Town.

In order to download a Web page, the Cape Town parts wholesaler dials their local ISP, then links through the ISP's Border Gateway Router into the Cape Town telephone exchange, then through standard telephone lines, up the coast of Africa, over the straight of Gibralter, through Spain and France, and then across the channel to England. As it does this data will inevitably pass through a number of routers and exchanges before it reaches your company's data 'caching server' in London.

The data 'caching server' picks up the slack and feeds basic web site data back to the parts wholesaler. This system speeds web page download because data comes from London -- not the west coast of the US. Only when needed, the 'caching server' icalls for specific information (such as part availability and part location) from your company's main web server in Silicon Valley. For the sake of arguement, we'll say our network operator owns a dedicated OC 12 line direct to a network sub-station in New York. In this case the web server in the Silicon Valley sends data, direct in 'one hop,' straight from the west coast to the east coast. In New York, the network Border Gateway Router 'routes' the new data back to the 'Caching Server' in England, which in turn sends the new data down to South Africa.

In fact for our data to make this trip, back and forth, it might cross ten or more nodes and will travel through a multitude of telecommunications cables. At the end of each leg of its trip it will find a new router that will read the IP information and then forward the signal down the chain of nodes. Each node consumes time, as the routers decide, and define, the shortest route to the destination point. Even having access to a dedicated 'one hop' Fiber Optic Cable is no guarantee of fast data transfer.

All kinds of factors will impact on the delivery of data directly to, and from, the server that hosts your site. It is not just the bandwidth of the switches and routers that connect the NOC into the Internet, but also the volume of traffic that moves directly into the NOC that has an impact on network speed. For instance an OC 3 line operating at 50% capacity may be as fast (or faster) than an OC 12 line operating at capacity. In the case of a networks with a large number of shared (or virtual) Web servers, the competition for network data paths can be even more intense because some companies load as many as five hundred sites onto one server.

Again it is bandwidth availability not capability, which is the ultimate register on which network value should be judged. A small network provider may still offer faster overall service in the end, no matter how big their popular competition is. Just as the rarely used country roads, which our archaeologist uses to travel to her 'dig' after getting her Land Rover fixed, may be faster than the highly trafficked multilane highways, which the network operator takes when heading into San Francisco after work.