Trilight Zone

[tricore]

Looking at things in a simple and naïve way, all one needs is a Web site, and anyone can get at your content. But, of course, your server's Internet connection may be inadequate to the task. Some users many routing hops away from you may experience dreadful performance, even when the server is not especially loaded.

For this reason, commercial content delivery networks, Akamai being the most famous, were established. They improve performance of a site by caching content closer to users, and provide security against distributed attacks. A network like Akamai's makes content delivery seamless to the user, who doesn't have to do anything special to access the content: it just comes through better and faster. And standards are developing like ESI (Edge Side Includes), which make the process of pushing the assembly and delivery of content to the edge of the Internet, closer to the user, more straightforward.

Free alternatives

This kind of service doesn't come cheap though. That's why free alternatives have been available for years. The general approach to doing a content delivery network for free is to build a peer-to-peer network, generally on volunteer sites, most of them some guy's PC in his home.

There are literally dozens of such networks, many of them developed for underground, nefarious purposes, such as illicit sharing of copyrighted content in a manner difficult for authorities to detect and shut down. Perhaps the most famous of these networks is Kazaa, which attained a significant installed base through bundling agreements. Kazaa has a shady reputation for installing adware and spyware, and has been plagued by legal problems over the content shared on its network. These problems are shared by many other players, including Grokster. Others, such as the very popular eDonkey2000 network, have fared better.

Founded much earlier, Gnutella is a protocol for distributed file sharing that has been implemented in numerous programs, including LimeWire, BearShare, and Morpheus. The original design of Gnutella was heavily influenced by the goal of privacy and anonymity for users on it, with the result that the protocol was inefficient and the network unscalable. Some changes have been made to address this, and a schism developed with the Gnutella2 protocol, but Gnutella still routes many operations through a circuitous node chain that dooms it to sub-optimality.

Much more technically interesting for large file distribution is BitTorrent. This protocol breaks the file into smaller pieces that can be distributed from numerous peers in the network. It has the downside of being relatively difficult to work with; a special torrent file must be created in order to share a data file, and there's no standard way to advertise torrents to other users. There are many BitTorrent client programs and some, such as Snakebite are beginning to address these issues.

Difficult to police

All of the open protocol-based file sharing systems share the problem of security. Since their goal of being resistant to shut-downs means there is no central authority running them, they are difficult to police. Malware is not uncommon on these networks, and many Internet worms are specifically designed to spread using file-sharing networks.

Another technically interesting method with a little more control, closer to the Akamai model in some ways, is the Coral Content Distribution Network. In this network one simply prepends the domain portion of any URL to that of the Coral server with 8080 as a port, for instance. Files with such URLs are referred to as "Coralized." Coral uses an indexing architecture called DSHT (distributed sloppy hash table), which helps to move content closer to endpoints.

It's difficult to envision a circumstance where the current generation of networks would be worth using for legitimate businesses, although perhaps some distribution of large media files through BitTorrent would seem cool to a limited set of younger users. Businesses are better off using services designed for them.