Streaming Forum Preview: Next Generation Content Delivery Architectures
The world of content delivery and content delivery networks (CDNs) is about to undergo radical changes, and I'll be hosting a panel called "Next Generation Content Delivery Architectures" at this year's Streaming Forum in London on June 25.
As many of our readers know, I remain close to the evolution of the CDN industry after many years in the space. While I continue to be heavily involved in several of the world's leading conferences in the space (CDN World Forum and TV Connect), I have felt that the sector has become somewhat caught up in simple management of the commoditization of a small number of platform-as-a-service (PaaS) models, and for several years has not been particularly evolutionary at a technical layer.
Yes, there was some innovation and discussion about CDN federation models, and there have been some schisms as we have moved formats and delivery models, such as the step from one session-based streaming protocol to another, and then the eventual abandonment of such models for HTTP.
Some may think that HTTP was the inexorable end of the road, and now the future is all defined within simple HTTP distribution.
The exciting thing is that is all about to get washed away—and fast. There is, finally, a really significant step change coming to not only CDN models but network infrastructure as a whole.
Originally telcos were built by networking many highly-proprietary appliances. One of the aspirations of the internet’s founding fathers was to make these markets for proprietary network technology less vendor-dependent (i.e. more competitive).
As we now see, the inexorable move to "commodity tin" in the world's data centers is proof that the power of today’s x86 processor is sufficient to run most tasks—particularly if the architecture of the application is designed for the distribution over such resources, and these resources are abundant.
So in the past five years or so we have seen a general trend toward virtualization where the task of the virtual machine is designed to work broadly on just about any tin.
CDNs are in no way new in this space. However, nearly all of them are one of two architectures:
- PaaS providers, which use a mixture of public cloud, their own hosting, and perhaps some third-party hosting, and abstract the client from the billing of these underlying service to provide a consolidated bill for the delivery of the service.
- IaaS (infrastructure as a s service) providers, which allow for the building of your own PaaS model by deploying virtual machines (VMs) within the IaaS environment, and deliver directly from your VMs.
Nothing so far will be new to readers. But despite the flexibility that VMs bring, they are still "units" that need to to be deployed with some careful planning. They can scale up and down in minutes, and many of today’s media workflows are developed with this degree of flexibility in mind. Workflows can take a few days of planning, and can only scale up and down trailing the demand by a few minutes if not tens of minutes. While this is a step change from having to manually reinstall proprietary appliances, it is just the beginning, and it is already becoming old news.
The big game-changer that is coming over the next three years is an effect of the emergence of the Kernal-based Virtual Machines (KVMs) These can be thought of as incredibly small virtual machines that can boot in a few milliseconds, use little resources, and run as if they were native. There is no hypervisor abstraction from the underlying metal—the KVM uses the OS directly on the physical tin—which provides direct performance benefits.
Conjecturing a little, you can build a PaaS network model to serve just a few packets of video, then tear it all down while the user pauses their stream—freeing up the resources for other clients /purposes—and then instantiate the entire distribution network as they press play.
On a single-user basis it may seem inefficient to bother to free up the resources for a pause, but when you have 3 million users all acting in this way, the difference in your infrastructure requirements is absolutely immense.
Significantly, for an operator, this means a dramatic reduction in their underutilization, allowing greater availability for redundancy or resale to other clients (etc.).
This evolution will enable telcos to deploy CDN in response to an individual service request, or to rapidly deploy massive database processing services and a countless number of other tasks—all with a few milliseconds' change time.
My own company’s media workflow software has been developed for delivery of end-to-end solutions for video and audio media workflows in this type of environment. It is my pleasure to be asked by Streaming Media if I could form a panel at the show, and I've invited a group of peers in the sector—all of whom I have had lengthy discussion with about this emerging evolution of the content delivery network. I'll be joined by, Richard Cooper, controller online technology, BBC; Steve Miller-Jones, senior product manager EMEA and APAC, Limelight Networks; Chris Swan, CTO, CohesiveFT; and Andrew Parker, CTO, multi-platform distribution, Cisco Systems.
I am deeply excited to throw this topic to both them and to the room, and I hope you will join us for what promises to be the first time in a while you have gone "Oooh, now that was eye opening."
The 2014 Streaming Forum will be held at the Park Plaza Victoria in London on 24-25 June. The "Next Generation Content Delivery Architectures" panel will be at 1:30 p.m. on Wednesday, 25 June.