GigaOM has a great article out now about the Top 50 Cloud Innovators. It’s a good read, but I wanted to take a step back and talk a bit about the theory of clouds and how they may work in the future.
Recently, the head of Google’s Infrastructure team, Urs Hölzle, held an interview in which he discussed the ideas behind the cloud and the differences between the cloud and a network like the Public Switched Telephone Network.
Urs conclusion is that the fundamental nature of clouds is inherently different than that of a network like the PSTN. Clouds are inherently iterative entities, places where the very basis of computing is being questioned on an almost hourly basis. That is in direct contrast to the PSTN which may not have seen an upgrade in the last forty years.
Highly iterative systems will have downtime. When the emphasis is on additional features, stability becomes a secondary point. While this is a bit of a cop out; the intention, I imagine, through it all is to produce excellent stable code but Google is, in particular, willing to sacrifice stability for progression. In Google’s case this makes a lot of sense, which brings me back to my central argument.
Like FaceBook’s open compute project, the future of the cloud is the future of computing imagination. We’re seeing a more specialized world every day, where even slices of individuals computer cores are used for certain projects. This is part of the cyclic nature of invention, first we will experience a period of specialization, a breakthrough will occur and we will experience a period of generalization. Right now, we are getting heavy into the specialization, and the world is ripe for disruption.
Of these top 50 Cloud Innovators, I am worried about the massive CDNs, which will prosper short term, but may fizzle over the course of the coming decades. In particular Akamai and Cisco seem to be very tied up in the packet tagging game. As the world becomes more mobile, CDNs will become more numerous and we will reach a point of saturation, sooner than we all think. The demand for video content is increasing faster than the capabilities of current networking hardware.
Something has got to give.