What Is a Content Delivery Network (CDN)?
This is an installment in our ongoing series of "What Is...?" articles, designed to offer definitions, history, and context around significant terms and issues in the online video industry.
One of the basic terms in streaming and online video is the CDN, which stands for content delivery network. But what is a CDN, exactly? CDNs are used to rapidly and cost-effectively deliver a variety of content to numerous end points, whether the end points are web browsers, mobile devices, set-top boxes, or even gaming consoles. This article will define CDNs, look at their history, assess the current market landscape, and finally discuss what CDNs mean to you.
CDN: Definitions & History
There's not really a generic CDN, but for our purposes, they fall into three basic categories: those that deliver general content including, but not constrained to, video delivery; those that deliver on-demand or pre-recorded/pre-encoded video; and those that deliver live video.
Let's take a look at the differences and similarities between each of these. All the technologies described below, from full download to progressive download to streaming to HTTP delivery, are in use today—in a variety of business models—although some have proven more popular than others.
General Purpose CDNs
The growth of CDNs started before video, and chances are you've used their services if you've downloaded a software update, bought a song on the iTunes Store, or even just visited a popular website.
These general-purpose CDNs perform what's most frequently referred to as web acceleration. This is generally accomplished best by a CDN that has a number of servers in many locations, ideally close to the large connection points between internet service providers (ISPs) or even within the same data centers as a popular website or gaming/application provider. The CDN caches content, storing a copy of content that will frequently be requested by a high number of internet users.
To understand web acceleration, think about it in terms of the marketing pitch that used to accompany those CDs we got in the mail for a certain dial-up services. The marketing pitch offered potential users the ability to "turbo boost" of their web browsing experience, at much faster speeds than their 56Kpbs modem would allow.
Accomplishing this was fairly simple, since every internet-bound user of the dial-up service provider needed to dial in to one of its points of presence (POPs). When a particular piece of content was request multiple times, the service provider would cache that content at one, several, or all of its POPs, meaning that any request for that content would never go out across the costly internet network lines but would instead be served up from within the service provider's own data cache.
Today, however, web acceleration is a bit more complex, as even the most popular ISP in the United States only provides internet connectivity to only about 22% of the total U.S. online population. The CDN delivering content that can be web accelerated needs to cache content at many locations, resulting in a need for both a national footprint as well as a large quantity of servers at each the majority of US data centers and ISP headends / POPs.
The general-purpose CDN market is highly fragmented outside of the United States, falling along language lines and country boundaries, except in a few parts of Europe and Asia. Even in those markets, the language clusters in a single country can lead to multiple CDNs being able to make inroads into various parts of the market. India is a good example of this, with more than thirteen different distinct official languages, depending on the given state or municipality.
On-Demand Video CDNs
Some general-purpose CDNs also provide CDN services for on-demand video content. The thinking is that video content is just a large file, like a game or large application download, so the serving of video content should not be much different than other content.
A few years ago, the disconnect between general-purpose and on-demand-video CDNs was quite distinct, as video delivery required the use of a streaming server.
Streaming servers deliver the content at the time of a request, but only deliver the bits requested rather than the entire length of the video clip. This was helpful for the content owner who was paying the CDN for delivery by the bit, since a viewer choosing to abandon viewing of a clip halfway through its duration would not have downloaded the entire clip, regardless of the viewer's internet connection speed.
CDNs didn't necessarily want to spend the extra dollars for hardware and streaming server software, just to limit the number of bits being delivered to the viewer, so a few other ideas were tried.
One was the direct download, where the entire video clip needed to be downloaded to a viewer's computer before it could be viewed. While this is acceptable for downloading a game or computer application, the tolerance level of viewers waiting for content to download—especially something such as a movie, which could be over 1GB in size—was low. No one wanted to wait, upon selecting a movie to view, twenty minutes or more to view the content.
Another option was the progressive download, which is used by a number of online video sites, such as YouTube. In this version, the CDN begins delivering the download, but a viewer can begin watching content within the first 3-5 seconds, on the assumption that the viewer's fast internet connection will continue to download the video clip at a fast enough pace to avoid running out of viewable content. As the speed of internet connections outpaced the bitrates used for standard-definition content, though, many viewers would have the entire video clip downloaded before they were even halfway through viewing the content.
Abandoning the video clip partway through still meant that the content owner was paying for full delivery, since the entire clip had been downloaded. This issue, of course, is exactly what a streaming server was designed to avoid, so the hunt was on for a progressive download solution that could throttle content delivery (curtail the speed at which content was downloaded to a level just a bit faster than content would be viewed) while still avoiding the use of a specialized streaming server for on-demand content.
In recent years, the best solution to emerge is called HTTP streaming, and it is coupled very tightly with adaptive bitrate (ABR) encoding and delivery.
Just like it sounds, HTTP streaming uses generic HTTP servers (often based on Apache or Windows Server) to deliver on-demand video files the same way that other HTTP website content—such as images and text files—is delivered. HTTP delivery has had the capability of throttling for a number of years, even before the advent of online video, and these same techniques can be applied to throttle on-demand video clip delivery.
The most interesting addition to the CDN arsenal is adaptive bitrate (ABR), which converts a video stream into fragments or chunks, often 2-10 seconds in length. ABR creates discrete streams at various bitrates and then uses feedback from the internet user's video player to dynamically detect the optimum network speed for delivery of the video clip.
As network conditions change – for better or worse – the stream with the most appropriate bitrate is served for that give chunk of time. As this is a streaming solution, it keeps the content owner happy by limiting the bitrate and delivery of content to just what the viewer will consume, and it benefits the CDN by limiting the number of specialized streaming servers needed for on-demand video delivery.
Not surprisingly, given the benefits of ABR, there are multiple ABR solutions in the marketplace today, including one each from the Big Three – Adobe, Apple, Microsoft – as well as several proprietary solutions, such as Move Networks. All act in a fairly similar manner, and some even allow delivery via HTTP, either HTTP streaming or progressive download.
Live Video CDNs
Despite the advent of ABR and HTTP streaming, there still is a need for live video delivery, since live video can't be cached like content that's previously recorded.
This area is probably the least mature of all the CDN models, for three reasons.
First, the vast majority of video content delivered by CDNs is on-demand video: some estimates are as high as 95% of all online video being delivered as on-demand video content.
Second, since live video can't be cached, it's necessary to modify the basic CDN infrastructure to have either very high-bandwidth pipes between a central location and the end-user viewing the content or to have slightly lower-bandwidth pipes that send the live stream to a repeater or reflector that is nearer to the end user.
Third, given the two points above, the cost to build out and maintain a live streaming solution for very popular live events is daunting: building and maintaining a million-plus viewer live streaming solution is quite expensive.
Given the fact that actual streaming servers must be used and that bandwidth delivery costs rise dramatically for each additional simultaneous viewer (on-demand content delivery typically sees simultaneous viewer bandwidth at one-tenth the number of live viewers) the live video delivery solution would need to be in almost constant use at its peak viewing level to reach the economy of scale of over-the-air television and traditional cable live video transmission.
That's not to say that live video CDNs aren't viable: the move by over-the-air broadcasters to use part of their digital frequency spectrum to deliver video to mobile televisions or handsets is just another version of the CDN, albeit one that has a sixty-year head start in terms of placing transmission towers and covering wide geographic areas.
The problem with the blockbuster online live event, however, is that it often takes a number of months of planning to increase the size of a live video delivery network to the level of anticipated audience viewership. If successful, and more viewers attempt to watch than the system is built for, there is a risk of technical failure; if fewer viewers choose to watch the live event, there's the risk of financial failure from overbuilding the CDN for a specific event. The challenge is great, but the industry has shown it can overcome large challenges, so expect to see significant movement in this part of the CDN space over the next two years.
Current CDN Market Landscape
Dan Rayburn, who tracks the CDN market for StreamingMedia.com as well as his own blog, has seen the market go from 27 vendors in 2007 to almost twice that many at the end of 2010. One reason for this is that the definition of CDN—as viewed through a streaming lens—is also expanding to the more “traditional” CDN services of small-file delivery and web acceleration. These aren’t new services, as the Tier 1 service providers has offered webpage acceleration for years, but their importance to streaming is growing with the continued inroads of Adaptive Bitrate and HTTP streaming.
Vendors in the CDN space often span the three areas mentioned above, so the list below breaks out CDNs by Tier 1 service providers, data service providers, and a few key stand-alone CDNs. It is not intended to be comprehensive; a complete list of CDNs is provided on Rayburn's blog.
Tier 1 Service Providers
AT&T, Bell Canada, Deutsche Telekom, Global Crossing, Level 3, Tata Communications, and Verizon
Data Service Providers That Also Provide Some CDN Services
Amazon and Rackspace
Akamai, EdgeCast Networks, Highwinds, Limelight, Mirror Image, iStreamPlanet, Octoshape, PowerStream, StreamGuys, Streamzilla
Technology Providers to Many CDNs
Adobe, Cisco, Jet-Stream, Juniper, Microsoft (Microsoft also has Windows Azure, a cloud computing infrastructure)
Why CDNs Matter to You
So why should you care about the history and current state of the content delivery market? The answer depends on whether you are a consumer, a content publisher, or a technology service provider.
For technology service providers, the need to stay current with trends in the digital media delivery space is key to offering services to content owners and CDNs. The trend toward HTTP-based video delivery may appear to be a “dumbing down” of delivery options, but as great amounts of premium content begin to move online, the owners of this content will be looking for ways to both monetize and protect content, while maintaining a lower overall delivery cost. All three of those needs converge at the CDN.
For consumers, an interest in CDNs may appear to be about as fun as watching paint dry, or understanding the inner workings of a radio or television transmission system. Yet the ease at which consumers can view content on any device at any time—one of the potential benefits of online video consumption—is directly tied to the advances in delivery and content rights management. Knowing what CDN delivers which content type makes you better informed when it comes time to distribute your own video clips.
For content owners, understanding the various types of CDNs, as well as the ancillary services such as web acceleration and application delivery, is beneficial when it comes to presenting your compelling premium content in innovative ways. Whether it’s reduced delivery cost for content not destined for the television screen, or repurposing content that can reach an audience much bigger than just a national market, content owners need to be knowledgeable about the trends in content delivery to guarantee protected, efficient forms of media delivery for the most cost-effective monetization strategies.
A high-level view of state of streaming media in 2019, and the concepts and and technologies that make it work.
Launching a CDN is becoming a more attractive option for many companies as a way to control costs and provide additional services.
The move to connected devices for entertainment options will happen faster than many would have predicted, claims a strategist from Akamai.
A high-level view of streaming media technology, history, and the online video market landscape