The Internet is poised on the edge of a content explosion. For years now there has been a group of companies that grew up to help with Internet content delivery with products called (no surprises here either) content delivery networks.
So, with content ready to explode, the vendors already out there are in the catbird seat, right? Maybe not, because this might be one of those times when more means less.
The growth of the Internet has created numerous technical challenges. One in particular is how to make rich Web experiences work for millions of users without terabit core networks and server farms the size of metropolitan areas being required.
CDNs developed along with rich Web experiences as a way of providing a shortcut for delivering pages of content or media files from a cache server closer to the user. The user makes a content request by clicking on a link, and the request pulls the content from the nearest cache instead of dragging it out of the Web site’s data centre and through a bunch of intermediary ISPs.
If CDNs are a good idea, it’s sure tempting to think that more content would make them a better one. If you look at CDNs’ value proposition, however, you can see some cracks. At the highest level, CDNs’ value lies in their ability to navigate the zone between so much content that caching is impossible and so little that caching is unnecessary. Surprisingly, the “more content for the Internet” trend is moving these extremes together and making that zone smaller.
Let’s suppose for a moment that the only IP content is television programming. In this scenario, every metropolitan area has its own IPTV server farm, because television programming is popular enough to justify serving it from each separate major metropolitan area. Under these conditions, the servers that CDNs are short-circuiting are in the same places the CDN caches are, so there’s no value. The network connecting the user to the CDN is the same one — broadband access and the metropolitan network — that connects the user to the IPTV servers. No bypass there either.
Now let’s suppose that, instead of IPTV, the YouTubes of the world own all the content and that millions or billions of users are uploading and downloading it happily hour by hour. What do CDNs cache? All the content in YouTube? Apart from the obvious storage challenge of keeping copies of the entire YouTube inventory in every cache, there’s the problem of how it can be updated there.
CDNs are based on the assumption that there are a manageable number of relatively static content elements that are accessed a lot. Under these conditions, the content owners might well find it better and even cheaper to cache the stuff via a CDN than to build their own metropolitan servers to host it for high-performance user access. YouTube-like content isn’t a good CDN target, and neither is IPTV content. That means that a shift to either type of content would reduce the benefit of CDNs.
Technology changes also are affecting CDNs. More peer-to-peer cached content, for example, could be delivered through peer networks, and some vendors are looking at doing that. More bandwidth is another example, because if core bits are cheap enough there’s no reason not to stream content from the source and forget the cache.
Because bandwidth is required to make an IPTV or YouTube experience work well, the content shift we’ve talked about makes it likely that more core capacity will be deployed, and that reduces the performance benefit of CDNs, even for nonvideo applications. Network performance may not be as much of a limitation in the future.
The increase in the number of broadband users poses a potential risk. CDN vendors get a big chunk of their revenue from such corporate giants as Microsoft. What happens if the richness of the experiences induces Microsoft to have their own metropolitan centres? Interest by players like Microsoft in content as a business makes this more likely anyway. If you’re delivering IPTV or other content from a metropolitan centre, why not do your own caching there?
This doesn’t mean that CDNs are going away, or even that they won’t be growing over time. Rich media on Web sites doesn’t necessarily mean highly dynamic content or full-resolution high-definition TV video. Even without these extremes, however, we’ve still got to wonder what the content trend will do to caching. Is more personalized advertising a good thing or a bad thing for CDNs? Will Web sites have to create more variable, interesting, TV- or YouTube-like pages to keep users viewing their pages from slipping into a coma? CDN providers certainly have plans under way to try to cash in (no pun intended) on personalization, advertising and other key trends. Will they work?
We always hear that content changes everything. It certainly appears the content dynamic will have a far-reaching impact on the Internet and on the companies that are supporting it as it evolves.