As the big guys such as Cisco make their play for the video over IP market, it raises the perennial question of return on investment.
Video is a demanding medium on technology, making simple web pages and images pale into comparison. As a result there's tendency to believe that high end equipment is necessary to build a content delivery network. There's a lot of truth in this argument, but, given a choice I would always go the 'more and cheaper' route.
A single Wintel box can cope with around 500 concurrent users, but a scaled up solution that can handle 2,000 connections would cost well over four times the cost of a basic box. And having more boxes provides more scalability, redundancy and flexibility. The only downside is the increased licensing costs for the server software. So, it's a Cisco v Microsoft trade off.
Inevitably, this technology goes out of date faster than a Playstation 3 at Christmas, so a large scale investment is also unjustified from an ROI perspective.
Scaling elements like firewalls become a very expensive undertaking and protecting individual smaller machines is more competitive.
But at the end of the day, with only a handful of providers really offering a true video CDN service, and few still offering bespoke hardware and software, it shows how difficult the architecture is to get right. For most building their own isn't an option.