Many of the benefits of cloud computing are lost in translation as enterprises attempt to force the "new wine" of cloud's flexibility into the "old bottles" of traditional data centers. By running a cloud environment within one's data center, the full benefits of infinitely scalable and flexible infrastructure fade, as Amazon …
Sounds like more cloudy snake oil to me. Relationships with CDNs are not cheap, and there aren't that many customer facing businesses that could benefit from having a 25ms latency for an HTTP request as opposed to a 50ms one. So you pay this outfit to tell you which CDN to use, and then pay a series of CDNs to host your site in every part of the world... then to top it off, you have to rearchitect your web app to manage even more cache sets and route requests to each of them dynamically. I'm not saying it's a universally bad idea, just a very diminished marginal return.
Deal with CDNs at work quite a lot and ...
... having a solid relationship really matters a lot, particularly if you're slinging around video and audio. And what happens if you want to secure your content? Akamai offer great customer service and battle-hardened security. That's something Azure, Level 3, Cloud Front, Mirror Image, LimeLight and Edgecast just don't offer - their knowledge pool just seems poorer.
Be curious to know what other people's experiences are with delivering secure video via a CDN?
Re: Deal with CDNs at work quite a lot and ...
YMMV. Akamai is not so hot if you drop out of their edge servers. Then your performance drops *and* you get to pay for the privilege of Akamai filling up its cache network. Recently moved to Cotendo which, while not having anything like as many servers, has given lower latency and better reliability. You can imagine that I'm not best pleased that Akamai is buying Cotendo after having failed in the courts. Cotendo has the better topology: persistent connections and peering arrangements are more important than edge servers.
Regarding video - this is a very specialised area so what works for one setup may not work for another. Recently had a secure webcast that failed spectacularly with Akamai in Asia where there are national firewalls everywhere. Can't remember what we went with instead but it was a local operation.
In general CDNs focus heavily on North America, then Europe (less of a problem with so many providers and fibre) and the rest of the world is very much an afterthought.
Matt's column should get rebadged as the VC's whore where he pimps the Valley's most recent darlings; Cedexis isn't looking for funding by any chance? And he should get a style guide: "Here's out it works" should be followed by a colon not a full stop. Snake oil salesman with poor writing skills. What is the world coming to?
Cedexis' diagnostics is nothing new - Browsermob has offered something similar for a while - but it is slightly smelly because it co-opts the user's browser for network analysis without their permission. IMHO that should be opt-in only even if it is reasonably anonymous. The Browsermob standard service will give you a reasonable idea of how well your site is performing around the world - currently restricted to AWS but that is due to change this year.
Multi-CDN is the trend now...
Have a look at another French company (mon Dieu): CleverScale; they provide a multi-CDN management tool with log aggregation.
@Jeff11, yes, it seems a marginal advantage to benefit from having 25ms latency (for most of the Internet it is), but there's more about that: think offloading under variable loads. Old-school CDNs are much more efficient and easy to maintain than a distributed architecture on the Cloud.
Nevertheless: not the cheapest solution around.
Horses for courses
It's possible you'd get a much bigger variation between CDNs in some cases - CloudFront offers great performance in Europe and North America, from what I've seen, but falls down quite badly in Australia (most CDNs do: apparently certain very big ISPs there don't believe in peering) - so using Akamai rather than CloudFront could make a huge difference there - not 25ms, but "same continent" as opposed to "another hemisphere". (Maybe Singapore, which might not be so terrible for performance, but still quite a distance compared to being at a big in-country peering point.)
For corner cases like that, it might even make sense to go non-CDN - colocate your own machine down under, route your Australian traffic to it instead of CloudFront. (Akamai have servers everywhere - but won't talk to small customers, and even terminated the reseller they caught selling to small businesses; Rackspace are using Akamai bandwidth for their Cloud Files CDN, but only enabling the most basic static-content subset of features.)
Is there a particular reason this is
Is there a particular reason this is categorised under "virtualisation"?
Plus, what Jeff11 says... the addressable market isn't big (in terms of number of companies), and the number of those companies that care about this kind of thing but don't care enough to keep the market knowledge needed for a "good enough" (maybe not best/cheapest, but 'good enough') service in-house is even smaller.
Still, there'll be an IPO in it for someone somewhere.