Video over the Internet is a relatively new phenomenon and it represents a fundamental paradigm shift in video distribution and even the architecture of the Internet.  Video distribution migrated from a purely wireless medium (television) in the early 20th century to primarily a cable medium in the latter 20th century.  While cable and other Multichannel Video Programming Distributors (MVPD) still dominate video distribution in the early 21st century, we are beginning to see a new migration to the Internet.  This has created a number of economic and engineering challenges that the Internet is only beginning to resolve.

The big change in video over the Internet is its change from a broadcast format to an on demand format, and people value on demand content.  With broadcasting technology (whether analog or digital), the same video is delivered to thousands or millions of viewers.  That means the data is sent only once but received by many.  Video on the Internet is typically delivered using unicast where each viewer is delivered their own copy of the video which requires tremendous server and network resources.

For example; a High Definition (HD) video with a bitrate of 15 Megabit per second (Mbps) broadcast over a high powered radio transmit tower requires 15 Mbps of network capacity to reach thousands to millions of people.  However, a 2.25 Mbps YouTube on demand video stream delivered to 1000 people via unicast requires 2250 Mbps of network capacity.  So despite the fact that the YouTube video is a lower bitrate than the digital HD broadcast, each person is asking for a new copy of that video over the network rather than the same copy.

The Internet has adapted to the video challenge by shifting the majority of Internet traffic from a dozen Tier-1 network providers to the edges of the network, and this often means directly connecting the content providers to the broadband networks or very close to them.  This doesn’t mean that the traditional Tier-1 networks are losing traffic; they’re actually increasing traffic.  What’s happening is that the traffic going directly to the edges is simply growing much faster and dwarfing the tier-1 core of the Internet.

There are two reasons for this shift.  The first is cost because direct connectivity or “peering” is simply more efficient.  The core of the Internet has to handle all of the communications happening between the edges and it simply cannot handle all of the on demand unicast video.  The second reason is cost because if we can cut out the middle man (the tier-1 provider), we can substantially lower delivery costs.  Yet some advocates see this change in the Internet’s architecture as something that must be stopped when it is actually a superior economic and engineering model that must be protected.