By Mari Silbey
Posting in Design
What does it mean when only a handful of companies control a third of the traffic on the Internet?
Many people made fun of Alaska Senator Ted Stevens when, in 2006, he famously described the Internet as a "series of tubes." But how many people really understand it better, or are aware of how much Internet infrastructure has changed over the last five years?
A study presented last month shows that more than a third of all Internet traffic - and at times, nearly half of traffic - is being ferried around by a small group of companies. That finding could prove significant because the companies that control Internet traffic also have the ability to decide how traffic is managed. At issue are variables such as quality of service, transport speeds and pricing, and whether and how subsets of content may be prioritized for delivery.
Study co-author Craig Labovitz, who is CEO of DeepField Networks, and Internet networking consultancy, says the latest research is new evidence that the world's web infrastructure has undergone massive consolidation in recent years. In 2009, Labovitz helped conduct another study which concluded that roughly a third of all Internet traffic was being delivered by only 30 companies designated as "hyper giants." These companies included Limelight Networks, Facebook, Google, Microsoft, and YouTube, among others.
However, his recent research goes even a step further.
Labovitz showed in May at the Content Delivery Summit how content delivery networks (CDNs) alone now represent between 35% and 45% of web traffic on average. That's up from the 20% to 30% range in 2010. And for context, consider that less than half a dozen providers control the vast majority of the CDN market. Following the trail to its logical conclusion, that means a small handful of companies, maybe four or five, control probably a third of Internet traffic today.
Just like we have a finite number of companies with the scale to maintain large airlines, railroad networks, or utilities, we also have a small defined group with the ability to power web traffic. Some of these are content delivery networks - companies that optimize delivery of bits and bytes to speed up the performance of many different websites and web services. Others are big content companies like Netflix, or hosting companies like Rackspace.
Adam Thierer, a senior research fellow at George Mason University's Mercatus Center, and the former director of telecommunications studies at the Cato Institute, says this kind of consolidation is inevitable.
The economics of network industries are not those of the corner lemonade stand. We're never going to have hundreds or even dozens of companies competing to provide the sort of sophisticated and secure digital infrastructures we all increasingly demand.
However, just because consolidation may be inevitable doesn't mean we shouldn't monitor its progress. From a network neutrality perspective, there could be cause for concern in the future if there is increasing overlap between CDN companies and Internet backbone providers, or between network providers and content companies. In those cases there is the potential for conflict of interest because of the economic incentive to prioritize one type of Internet traffic over another. That's where consolidation could become a problem.
As for today, the Internet is already far denser in structure than it was originally conceived to be, or than it was a few short years ago. What will the Internet look like in another five years? If the trend persists, it will be concentrated into even fewer hands than it is today.
Image credit: Original network photo by Small_Realm
Jun 28, 2012
All of these big CDNs such as Facebook, Google (including Youtube), etc. wouldn't be such a large part of the internet if they weren't so popular. But it doesn't mean that less popular destinations haven't grown as well. Today you can go to university websites and download entire courses. Museums digitize whatever they can and put it on the internet. Consumers can download user manuals, engineers can download component specs, financial data is readily available, Supreme Court rulings are online, newspapers have archived old editions, etc. People can now do most of their banking, file their taxes, and do other essential tasks online. Back in the '90s when the internet first became popular, very little of this was possible. While none of these will ever be the most popular destinations (with the possible exception of Wikipedia), they all make our lives a lot easier at one time or another. I can't imagine these ever going away.
Consolidation implies a compression of available choices, but it is the opposite. Just look at that percentage of CDN traffic as a share of total internet traffic; while the top end increased, the bottom end is 3%. People are exercising a wider choice of destinations, some of it is within ecosystems, others are outside of it; depending upon the time of day, that share of traffic dramatically changes. If people stop using Facebook, Facebook's server content farms will suck less traffic. Facebook can't dictate to its users the speeds at which it can access Facebook servers -- to do so is to commit suicide. And CDNs were created to optimize speed and service, by reducing usage of backbone providers whose pipes can be willfully turned off (dark fiber) to maximize rents but also slowing down delivery.
That is true, but given that the major infrastructure companies (including CDNs) are building and integrating further services into their network offerings - moving up the stack, as it were - I believe the trend still shows a consolidation of power. SaaS companies that also own huge distributed delivery systems have a competitive advantage and create a barrier to entry for new market entrants. This is to a large degree inevitable, but it doesn't mean it shouldn't be monitored. There's also the issue of major backbone providers that want to separate out network capacity for CDN services. CDN services can make them money, but if they take away too much capacity from basic Internet services, that's another issue. It's the same argument I have with the ISPs who are now controlling both dedicated networks for IP video delivery and data networks for Internet service delivery. It's all well and good to separate out bandwidth for dedicated video delivery, but who decides how much capacity goes to those dedicated networks versus the "open" networks that subscribers also pay for? If there's not enough capacity given to Internet delivery, performance suffers. Competition should control for that, but only if there is enough competition to create effective pressure.