The State of the Stack
Players in and Around the Content Delivery Sector
Figure 1 shows the stack and a loose grouping showing where I think the various service provider groups offer services.
It's interesting, looking at how the horizontals relate-it clearly shows why IaaS clouds like EC2 are well set up to potentially take on not only the CDN but also the ISP (certainly when it comes to ISP hosting revenues).
Quick Roundup
Let's take a high-level view of some of the things of note that are emerging in the layers of the stack around the CDN space.
Application Layer
• IIS Smooth Streaming, Adobe dynamic streaming, HTML5/VP8, Wowza, and Flumotion are lining up to be the next-generation streaming format servers after Windows Media Server streaming andFlash RTMP streaming.
Presentation Layer
• Almost as soon as H.264 was settled on by Adobe and Microsoft, Google crashed the party with VP8 (presumably to avoid MPEG LA license fees on YouTube; one can only speculate that those fees would have cost Google more than did On2). If so, that will set an interesting tempo should Google launch a suite of video publishing tools, particularly if they are integrated with their own virtualised infrastructure and PaaS offerings.
• DRM remains relatively lacking in notable events beyond a few high-profile implementations. Laws are changing, though, and this is expected to bring in a drive for more DRM adoption and development. Popular sentiment is to pay a degree of lip service to the DRM; the content is often regularly purchased and then copied using commonly available tools. This area seems to be slowly working itself through policy courts while technology is becoming more and more sophisticated.
Session Layer
• Fragmented MPEG-4 containers that can be served from web servers have become vogue. The primary selling point is switching between streams of different bit rate compression in real time and seamlessly, so the quality can be downgraded and upgraded without interrupting viewing. Encoding tools are slow to market and expensive-particularly live encoders-and this is slowing uptake.
• The ability to serve data from widely available HTTP servers-but with the advantages of some of the flow control aspects of what was formerly limited to the domain of the media servers-represents a change point and probably the most interesting one, where the innovation cycle has perhaps ended. Despite more efficiency in less successful protocols and their respective media servers, in practical terms, modified Apache or IIS servers are commonplace and can easily now support not only more efficient file transfer but also live streaming.
Transport Layer
• Petabytes of transit and terabits per second of throughput and peering have becomecommonplace. This scale was unheard of 5 years ago.
• Prices across the CDN space continue to drop, as a few larger providers aggregate huge traffic and command immense buying power.
Network Layer
• Telcos have been on and off over the years but are hedging in favour of a direct play in the CDN space. KPN, BT, and AT&T all come to mind.
• New vendors are appearing, focussing on kits for CDNs; examples include Edgeware, Blackwave, Verivue, and Blue Coat.
Data Link
• Cisco and Juniper are accelerating their multicast deployment, and the routers arebecoming more and more functional for
video-specific purposes.
• HbbTV, Project Canvas, and Digital TV Group are all producing good standards to make IPTV and over-the-top video simpler for publishers to deliver popular content to a multitude of connected internet appliances. With this, over time, comes a new generation of core switch gear and routing.
Physical Layer
• Fatter and busier consumer networks need better and bigger core telcos. There must, at some point, be a natural levelling off, but while the internet is still growing rapidly, the content delivery market will require larger and larger pipes. As it does, it is gradually becoming a very important buying force in the telecoms market. In some ways this could be interpreted as a shift in interest for the telco from minute billing to packet billing. Indeed, many telcos are fascinated by packet billing for data transfer; VoIP was the first big shift, and now, perhaps, they are learning from the CDNs how to bill for video delivery on their networks.
Conclusion
A nuanced understanding of the stack really comes into play to help one see how all the various service providers in the sector interoperate, and what is noticeable is that the organisations that focus on their position in the stack are some of the strongest.
The terms "vertically integrated" and "horizontally integrated" are often used to refer to stack models. If one starts a streaming project with a clear picture of the stack, then vertically integrating and horizontally integrating all the suppliers can make the project as simple as building with Legos. By contrast, not having a picture of the stack can lead to booking multiple providers for the same job and omitting key providers for small but critical links in the infrastructure.
Companies and Suppliers Mentioned