The Internet is global in spirit. But the traffic associated with it is increasingly local in nature, as we found in our recent VNI study.

Consider a typical Internet user in North America. Today, over a third (35%) of the content accessed by this user is sent from a point within the same metropolitan area. Thanks to content delivery networks, a growing amount of Internet content is stored close to home. By 2021, over half (51%) of the average North American Internet user’s content will be delivered from within their own metro network.

The global numbers are lower, but the trend is the same. Globally, 22% of Internet traffic is delivered within the user’s own metro network today, and that will grow to 35% by 2021. Metro-delivered traffic will grow twice as fast (39% CAGR 2016-2021) as traffic traversing long-haul backbone links (19% CAGR 2016-2021).

There’s no mystery about the localization of Internet traffic: content delivery networks are the primary factor responsible for the trend. (The other factor is data exchange at the edge by IoT modules – a topic for another post, perhaps.) By 2021, 71% of Internet traffic will be delivered from a content delivery network (CDN), up from 52% today. CDNs are more developed in some parts of the world than others – in North America, 93% of Internet traffic will cross CDNs by 2021, while in the Middle East & Africa the CDN traffic share will only be 34%. Even where CDNs are well-developed, CDN traffic is not automatically local. Much CDN traffic still traverses regional core networks and even cross-country core links, either due to lack of presence in the end-users local area, or due to content routing that depends on factors other than distance.

In our VNI estimates, CDN traffic includes traffic both from third-party CDNs like Akamai and Limelight, and from private CDNs. Private CDNs are those operated by content providers for their own content. Google, Amazon, Microsoft, Facebook, and Netflix are examples of content providers with their own content delivery networks. Private CDNs are currently responsible for 61% of all CDN traffic, and this will rise to 68% by 2021.

These developments can raise a bit of angst in anyone with a fondness for the decentralized early days of the Internet. The content is concentrated in the hands of a few companies, and the delivery of this content may bypass much of the Internet’s infrastructure if it is delivered from within a user’s metro area and traverses only a single service provider’s network, so it isn’t “Internet traffic” in any meaningful sense. To an Internet veteran’s ears, this sounds more like a description of a broadcast network than a description of the Internet. But the decentralized and global spirit of the Internet persists in the discovery and exchange of content and information, even if not in the hosting and delivery.

Personally, I am not bothered by the prevalence of a few in content hosting or the role of CDNs in content delivery. It does seem to me, however, that replicating a broadcast-style infrastructure is costly on a global scale. Is there an alternative? There are some early indications that the intelligence possible with software-defined networks (SDNs) may help routing become so dynamic and adaptive that CDNs don’t need to become universal in order to make low-latency responsiveness a reality for most Internet users.

In the short-term, content delivery networks are here to stay and are likely to increase in importance with the advent of live video. VNI projects that live video is poised to grow 15-fold and will account for 13% of all Internet video by 2021. Live video will require low latency and high quality, and distributed delivery is ideal for live content such as sporting events and concerts that attract large audiences.

Video in all its forms will make up 82% of IP traffic by 2021, up from 73% today. The impact of video on the Internet has been transformative. Whenever the analysts on the VNI team are asked to explain a certain change in traffic patterns (the increase in peak to average ratios, the factors driving offload in mobile, the relative growth of upstream and downstream, to name a few), the answer usually has something to do with video. One of the most important but least visible implications of video has been the growth of CDNs and the subsequent reconfiguring the topology of Internet traffic.  Video, and the CDNs that deliver video, are bringing the Internet closer to home for billions of people around the world.

VNI Forecast Resources


Arielle Sumits

Senior Analyst

Service Provider Marketing