We’ve previously explained why data centre Tier may not tell you the whole story about a facility’s level of resilience. Much as the Uptime Institute’s classification scheme has been adopted by colocation providers and marketers, it’s not always a helpful guide for their buyers – particularly when two or more data centres within the same Tier classification level can vary wildly in terms of actual protection and strength of design.
What, then, should you be asking your colocation provider about resilience to avoid data centre downtime? Here are five questions we think are no less important than Tier classification, but are often overlooked by buyers who don’t know any better.
1. How diverse is the power distribution?
Many businesses don’t realise that two data centres can have the same Tier classification, yet offer completely different levels of power protection. Tier 3 in particular (which is a level claimed by the majority of commercial colocation facilities) states a requirement for at least N+1 UPS redundancy, but there’s still a lot of potential variation in terms of the diversity of other critical elements in the data centre’s power delivery and protection.
This means, for example, that one Tier 3 data centre may offer dual rack power from a single N+1 UPS bank on a single power distribution feed, whereas another may offer ‘real’ dual power to the rack with each feed delivered from diverse UPS banks, each being on an independent power distribution path for far better levels of resilience and protection.
A single UPS bank (even in N+1 mode) often means that the rack is connected to just one set of UPS modules, fed from a single string of power. If anything happens to one module within this UPS system, there’s enough capacity remaining for adequate cover. However, in the event of a complete UPS system failure, or a failure higher up the infrastructure chain, the single power distribution string means that both power feeds to the rack, protected by this single UPS bank, are at major risk – and so are all your servers and data.
In contrast, as can be found in our Manchester data centre, a diversely fed dual parallel pair configuration may feature two separate pairs of N+N UPS systems – with each pair fed by independent power strings – you take one feed to the rack from one pair and one feed from the other.
This is a big step up in terms of diversity, delivering UPS redundancy levels of 2(N+N) to your server rack – even if a catastrophic failure occurs at the changeover panel, main distribution or sub distribution level on one of the two diverse power feeds, or in the case of a complete failure of one of the UPS pairs, you still have fully protected power to your rack on the unaffected diverse infrastructure.
(Although meeting Tier4 level UPS configuration, this level of resilience doesn’t quite take you to Tier 4 level overall, though – that would require two independent feeds from the grid, which is usually prohibitively expensive to just about everyone apart from the largest enterprises, government facilities and critical national infrastructure sites.)
2. Where do you get your power from?
Moving up the chain, you may also want to ask your colocation provider where they actually get their power from. Data centre or otherwise, every building draws electricity from the grid in one of two ways – through a spur or a ring. A ring usually means you are part of the main regional grid supply, whereas a spur branches off from the ring.
Being fed on a ring offers more redundancy than a spur because mains power can travel to your site infrastructure from two independent substations – you can lose the supply on one side, but remain unaffected due to supply from the other side of the ring. Critical infrastructure sites are built on rings for this reason, as is our data centre – in fact, we share a Priority A ring with a local hospital, which means we’d be at the front of the queue for having our power restored in an outage.
Other data centres may be built on spurs, whether for reasons of cost or the simple availability of land in a built-up area.
3. How much time is on the UPS/backup generator?
If a failure within the mains or data centre infrastructure occurs (which can and does happen), you’ll be at the mercy of your chosen colocation provider’s UPS units and generator sets – and, specifically, how much time is available on their batteries and fuel supply to keep your business going. As such, enquiring about the data centre’s UPS and on-site fuel capacity is a must.
Be cautious, however, of providers who only provide you with a number and no indication of the other measures they take to prevent prolonged outages. Whilst 12 hours may seem like plenty of time to recover from a minor power outage, some data centre power failures can take weeks to fix if bespoke components are damaged and need to be replaced. That is, of course, unless the data centre has a backup part on-site – so make sure to look for evidence that the provider has other backup precautions in place, too.
4. How is the data centre cooled?
If infrastructure isn’t cooled within a data centre environment, it won’t last long, which is why most quality data centres make use of hot or cold aisle containment systems and high capacity units with a minimum of N+1 redundancy (which is required by Tier 3), thus ensuring that cooling is resilient against unit failure and working in the most efficient manner. This is crucial to guarantee that the different zones within the facility are cooled to an acceptable standard, and within SLA on a consistent basis.
However, you may want to ask your colocation provider questions about the quality of the air conditioning units themselves: are they low capacity, aging units that look like they might fail at any moment, or higher quality models with built-in redundancy via failover components and pipework? Don’t be afraid to ask for maintenance records either!
Again, with cooling, power distribution to the units comes into play – for example, as we have ‘real’ dual power supply at TeleData, we’re able to feed the cooling units from both the A and B sides of the power distribution for maximum resilience.
5. What is the management like?
Finally, even the most diverse, well-connected and sophisticated data centre won’t be resilient without a good management team holding it all together.
You’ll want to ask your colocation provider how they manage the facility on a day-to-day basis, including their maintenance practices and policies. There’s no use, for example, in having the latest cooling technology if the rack is going to fail because of a build-up of dirt and dust.
These 5 questions will reveal something much more than technical specification – they’ll tell you how deeply a data provider themselves understand their infrastructure, and whether they go the extra mile in building a truly resilient site.