The shift to the Internet video distribution presents a formidable engineering challenge for the Internet because of the massive bandwidth requirements of unicast video distribution.  Unlike broadcast communication technologies where data is transmitted once to many people, unicasting requires a new data transmission for each recipient.  So while a 15 megabit per second (Mbps) HD television show being broadcast to 100,000 people in a city only requires 15 Mbps of network capacity, a 2.25 Mbps YouTube video being unicast to 100,000 people over the Internet would require 225,000 Mbps which is a staggering amount of Internet capacity.  To overcome this technical challenge, caching and multicasting solutions can be employed.

Why caching and multicasting

On a technical level, unicast video is simply unscalable, especially at the core of the Internet.  While it might theoretically be possible to support 225,000 Mbps on some network links, we can’t even humor such an endeavor when there is 100 or 1000 times the traffic.  Even if it becomes possible to transmit that much data in the future, it will always be more efficient to use caching or multicasting techniques which allow for much higher quality and quantity content.  On an economic level, using the core of the Internet is cost prohibitive because long haul transit rates through the core are typically 3 times more expensive than paying for short haul peering at the edges.

Edge caching

The most common solution is to employ edge caching.  While this doesn’t change the fact that an aggregate of 225,000 Mbps of bandwidth is still needed on the “last mile” broadband portion of the Internet and many of the distribution networks near the last mile, it does alleviate traffic on the core of the Internet.  The bandwidth is also divided up into more manageable chunks near the edge of the network such that each caching server and the networks they support might only need to handle a maximum of 10,000 or 40,000 Mbps.

Caching solves a lot of problems but it doesn’t alleviate traffic on the last mile.  While that’s no problem for DSL or VDSL2 networks like AT&T and Qwest or for Fiber to the Home (FTTN) networks like Verizon FiOS, it becomes a problem for cable broadband networks and an even bigger problem for wireless networks.  Cable broadband networks are shared and wireless networks are shared as well as being more capacity constrained.  To a lesser extent, capacity constraints can also affect the distribution networks that feed the FTTN or VDSL2 networks.

Multicasting

Multicasting can resolve the last mile capacity problem because it is an analogue of broadcasting that operates on Internet Protocol (IP) networks.  Data is transmitted only once to all recipients on a network and each recipient receives the same data feed.  This can dramatically reduce the load on every part of the network including the distribution and access networks.

Note: MediaFlo is an example of digital broadcasting but it does not use IP and it requires dedicated radios or devices for viewing.  Cellphones that support MediaFlo require a second radio and second antenna.  A multicast solution on a wireless network would allow any generic Internet-enabled device to receive broadcasts over an IP network.

One downside to multicasting is that it may not be technically possible on every network and it does not work by default.  Networks have to collaborate with content providers to enable multicasting.  For example; the BBC works with many UK based ISPs to enable radio and television multicasting.

The other challenge is that multicasting may not be something the users desire.  More often than not, people don’t want to watch the same content and even if they want the same content, they may not want it at the same time.  Multicasting only works when a group of people want to watch the same content at the same time.  This doesn’t work too well for archived content, but it works great for live content such as sporting events or live news.

Multicasting might also work well for some software distribution tasks such as application updates, but this would require software on the computers that would have to remain operational, but software distribution doesn’t have to be sequentially delivered like video content and it can be delivered out-of-order.