Where did all the Tier I’s go?

It’s all about the cloud, baby.

At least, that seems to still be the perennial theme. You can’t thumbswipe your {insert preferred tech news source here} mobile app for another story about reducing cost and Carbon through cloud services. It’s 2016 and we’re still not bored of it, apparently.

Distributed data, global load balancing, virtualisation, containerisation, scale up, and scale out.

It’s all very, VERY, cool, and to be fair, compared to what in hosting terms is now seen as “traditional” managed physical hosting, cloud has reduced cost and improved time to market. It has led to the evolution of tech jobs from SysAdmins to DevOps. All good things, truth be told.

The issue however is that whilst we have chopped up compute and storage into little bits (literally) and can rent what we need by the hour, what sits underneath is invisible to us.

The megaproviders such as Amazon and Google tell us they achieve efficiency through designing their datacentres from the ground up. They use green energy sources, they use fresh air ventilation or evaporative cooling, they design their servers to take DC power supply directly (no pun intended) and have on-board batteries instead of centralised UPS systems.

For smaller entrants into the cloud market however, they rely on third party datacentres. For mid-sized and larger businesses who wish to build their own cloud platforms, they either have to build or rent from the same third party co-location providers.

To achieve similar absolute efficiencies however, this means thinking differently. You can’t keep buying HP or Dell and wonder why you can’t get your PUE down to 1.1

At the building level, equally, you cannot expect to match the price/performance of a scale out cloud provider who is using a huge shed with fans, when you are using a Tier III facility, with all the inefficiency of AC/DC conversion in the UPS, and traditional HVAC with in-row target temperatures of 25degC or less.

And if the whole point of cloud technologies, distributed data, and high availability software is that it doesn’t matter if your server (or datacentre) falls over, why do we still continue to buy equipment and facilities AS IF IT DOES MATTER?

The answer appears to be that there is no alternative.

Whilst there appears to be an element of oversupply in the datacentre market, there is limited actual choice. Almost everyone is providing Tier III (or the ever-fictitious Tier III+) quality facilities, and pricing pressure is driving margin and service erosion.

So the real question is, if at the OS, system and application layer everything is becoming increasingly “cloudy”, why are there no datacentres designed for this new paradigm?

In short, where are my cheap, energy efficient, cloud-ready Tier I datacentres?