Cloud Inside - New Generation
While the OVP space was increasingly focusing on client-side technologies, there was also a significant macro trend emerging in the Internet and telecom space: virtualization was becoming reliable and accessible, and General Purpose Commodity Compute was becoming powerful and yet affordable enough to be considered to replace the wide range of traditional application- specific highly optimized technology that had been used to that point.
With CDNs representing one of the largest, scaled-up, and most distributed service network applications, they were no stranger to this emergence, although right up until 2015 CDNs still thought they could deliver better service levels on appliances than on commodities off the shelf.
And what is interesting is on an individual machine for a specific task that may be true, but when you scale out the total cost of operations, and match that to the total availability of the resource, it is now almost invariably a better longterm strategy to fully virtualize.
The traditional CDNs are “schismed” about the topic, defending their “value proposition” over a public cloud with an endless list of KPI benefits, the fact is that the quality and pricing is good enough that many small publishers have built their own small delivery infrastructure in public clouds. While to a CDN an individual customer that revenues $100 per month is a client not worth investing in, the fact is that when that becomes hundreds of thousands of small clients who support themselves, the question has to be asked when is “good enough” good enough?
Some publishing companies have moved to cloud entirely. The BBC almost entirely delivers its iPlayer platform to the world from Amazon's EC2 service.
Conversely, CDNs and Telcos are increasingly changing their internal operations to leverage virtualization and cloud and to increase their service velocity (see Section 2.9).
So, while CDNs were themselves arguably the first really big SaaS operators, they themselves are conflicted about Infrastructure as a Service, and at the moment most teams in the CDNs offer considerable variations of viewpoints.
This will probably crystalize suddenly at some point in the next year or two, as one manages to leverage a high-profile release in a specific direction. But for now most CDNs say they are virtualizing only where it adds benefit and that they can't reveal anything, but yes, they can do everything you might ask them to, ..., etc.
In the next sections I am going to explain a little about what I think the CDNs should do, since if they don't, then the Telcos on which the CDNs exist are going to creep up the stack and disrupt the pure content delivery space in a way that, until now, the CDNs have traditionally felt dominant in.