Management consultant, Peter Drucker, famously once said; 'If you can't measure something, you can't manage it.'
This message has taken quite a while to reach the data centre/colocation industry, but there are encouraging signs that more and more data centre owners, operators and users are beginning to grasp the importance of understanding all aspects of both the facility and IT infrastructures. With this knowledge comes the ability not just to efficiently manage the existing data centre/colo facility but also to plan how to improve this environment and, ultimately, to optimise it.
Right now, there’s so much going on in the data centre that either needs measuring and/or monitoring and/or that can also be a part of the management solution. Take, for example, Big Data. The number crunching and subsequent analysis of large data sets requires a new approach to data centre design and operation – with some tricky decisions to be made as to where the Big Data operation needs to take place – at the edge, at a centralised location, or, most probably, a mixture of both. So, understanding the Big Data operation is key to ensure that it can then be provisioned and operates efficiently.
At the same time, more and more organisations are realising that the Big Data infrastructure that is helping to analyse customer data can also be used to analyse the effectiveness of these organisations’ facility and IT environments. Deciding when to use the Big Data application to analyse customer data, and when to use it to analyse ‘internal’ data is just one example of how intelligence in the data centre has an important role to play.
Similarly, as the promise of smart cities becomes more and more of a reality, the value of power in many countries across the globe (and, eventually, all, when energy prices are high everywhere!) will gain in importance. The cost of buying and using power at peak times will be substantially higher than off-peak power. So, data centre managers might like to understand more about when their customers need to run their applications, so when there is a high demand for power, and when there are quiet times. It may well be that data centre owners/operators can start to incentivise their customer base to consider how and when they use their applications/power. Plenty of customers might consider running applications at off-peak, likely night-time periods. Sounds fine – but then don’t many organisations back up their data during the quieter times?
And what about organisation that have multiple data centres across multiple countries? The different power prices being paid in different could well have an impact on where data and applications are held and/or run.
Facility and IT infrastructure refresh is a perennial data centre/colo topic. In an ideal world, kit would be almost constantly replaced, to ensure that the data centre is operating at maximum efficiency. Of course, the reality is that the cost of this constant refresh programme is likely rather higher than the presumed operational improvements so gained – not to mention the potential disruption caused by facility and IT hardware being installed, removed, installed and so on. The link between human intervention in the data centre, to solve Problem A, and the occurrence of Problem B shortly afterwards, is well understood. Humans cause problems in data centres!
So, of course, let’s remove the humans and replace them with robots. Um, yes, for simple tasks within the data centre, that’s possible, and in time many more operations and maintenance tasks could well be carried out by robots, but will they ever be able to carry out some of the incredibly delicate and/or awkward maintenance and installation tasks? Re-think and re-design the data centre, so that there’s a bit more space in which to move around, and easier access to the hardware in the data centre, and maybe robots can do it all?
However, for the time being, artificial intelligence, machine learning, neural networks and the like are better used in the measurement, monitoring and management of the data centre. In simple terms, every piece of kit inside the data centre – whether it’s related to the facilities or the IT side – can be reporting back to the centralised operations centre – providing details of the power it’s consuming, the work it is undertaking, its ‘health’ and a whole host of other parameters. Orchestration, or management software can, with or without human intervention, then attempt to optimise the whole data centre environment.
One thing to bear in mind – whatever the level of automation achieved in the data centre, there will always be the vagaries of human beings with which to contend. Last thing Friday night, the following week’s workloads are input into the orchestration machine. A few minutes later, and the machine has calculated exactly which applications will run, at which time, and on which servers, over which networks, and linked to which storage. Everyone goes home for the weekend. Monday morning comes and, oh dear, the MD forgot to ask for a major design and quotation to be factored in to the week’s calculations…the orchestration machine has to do some more sums. And then the marketing department wants to fire up a quick, web-based promotion campaign by lunch time…
Flexibility and agility are seen as something of a data centre/IT nirvana, but the reality is that an infinite degree of flexibility or agility means endless recalculations, and significant wasted time. Or, one builds in a degree of potential change into the weekly sums, but if there are no variations, then inefficiency has crept in!
That means that, at some stage in the automation process, either a human being and/or a piece of software has to decide just how granular the data centre optimisation process can, or needs to, be.
How smart or intelligent can a data centre or colo facility become? This questions is best answered with another question: “How smart or intelligent do you wish your facility to become?”
Only by answering this, having talked to the data centre folks, the IT folks, the finance folks, the marketing folks and the board, can you then embark on the desired level of ‘smartification’.