4D's CTO, David Barker, talks about the different data centre tier levels, what they cover and whether they're still relevant.

Transcription of the video is below:

"Today we're going to be talking about the different data centre tier levels, what they cover and whether they're still relevant.

The data centre tier levels were first started by the Telecoms Industry Association to define four levels of reliability within data centres, and how well they could survive a fault such as a mains failure or cooling failure.

A few years later, the Uptime Institute brought out their version of the tier standards, which covered four levels of reliability, but they also offered a certification option, and even today there's confusion between TIA and Uptime Institute standards. 

We'll focus on the Uptime Institute's tiering because that's the one that most people use. The tiers run through one to four, with one being the least redundant and four being the most resilient.

Tier One has little to no redundancy and if the power goes, the whole site will go out. There'll be no back-up power generation and it's only useful for applications and systems that you're happy to go offline at pretty much any point.

Tier Two, while it may still have a single path for power and cooling, there will be back-up generation systems and some level of redundancy built-in. 

Tier three introduces the concept of concurrent maintainability. This means that you can lose any one component, or take one down for maintenance, while still providing service through to the data floor and the applications.

You may have a slightly lower level of redundancy while these components are out but you will have failover and resilience in there. So you'll typically have a couple of generators, additional air-con units (so you can take a unit off for maintenance), and some additional UPS. Tier Three is the most common standard within the UK which data centres align themselves to.

Tier Four is the top tier within the Uptime Institute's levels of tiering and, with that, you're looking at full resilience throughout the entire site. So you'll have two completely separate power paths, two completely separate cooling paths, two completely separate network paths and, generally, more than one feed off the mains grid, and this means that you can lose any component within the data centre and you'll still have full resilience out to the data floor. Each of those power paths will have its own resilience in so you can think of it like having two Tier Three data centres because you'll have Tier Three standard on each of those paths.

There are very few Tier Four standard data centres, just because the cost of adding in that additional level of resilience generally outweighs the benefits of doing things like good proactive maintenance and good testing of the systems. What you should really do is go and do your own due diligence. So go visit the facility and meet the data centre team that is running it.

Find out what kind of planned preventative maintenance they do, find out what testing they've put in place, and what kind of monitoring they're doing.

If they're doing things like full building black-out tests, where a mains failure is simulated, that's really good for ensuring the generators are run properly and the UPS's are working effectively.

Having a well run Tier Three data centre will generally give you much better uptime than a poorly maintained and poorly run Tier Four data centre. Another good thing to look at is the overall level of resilience and uptime you need in your systems and applications. Just having your system in a Tier Three or Tier Four data centre isn't going to guarantee you a certain level of uptime and you should consider splitting your system, or your applications, across multiple Tier Three data centres - that way you can handle your failover and your resilience in the application layer and you're not so reliant on the data centre itself.

Even a certified Tier Four data centre is only certified to 99.99% uptime, which still means you can have up to 29 minutes of unplanned downtime per year. So an application split across two Tier Three data centres will give a much higher aggregate uptime than having it hosted in a single Tier Four site.

I hope that was useful and provided a bit of background to the tier standards you may have heard of when looking at data centres, as well as providing a bit of real-world context as to how you should approach them.

You can contact us at 4D to find out more about us and arrange a data centre tour."