One day, all data centres will be made this way

A visit to a modern data centre is a reminder of how on-premise IT has some way to go to match up to power of cloud

Every now and then it’s worth being re-acquainted with what this cloud stuff is all about. After all, there is so much marketing flying around that one can lose touch with the technology that underlies cloud computing – to the extent that it becomes little more than a few coloured squares on a Powerpoint slide.

A couple of weeks ago, I joined a number of journalists on a visit to two Equinix data centres in Paris (one complete, one under construction), on the invitation of one of their prime tenants, BSO Network Solutions. Now that the air conditioning noise has stopped ringing in my ears, what did I learn?

First, that it’s still about computers, storage and networking. I know this is an obvious thing to say; nether do I underestimate the amount of engineering effort that goes into every layer of technology, from chips to servers and storage devices. Indeed, one could not fail to be impressed by the rows upon rows of rack-mounted devices, each being maintained at a constant temperature using the latest in cold-aisle cooling.

But all the same, computers are still computers, even if they are now powerful enough to run several virtual machines in parallel. And an MP3 file will look exactly the same whether it is stored on a 50-terabyte array or a memory card. Whether we’re talking about the largest server or the processing capabilities of a smart phone, ultimately what matters is how tasks are allocated to which processors, and how information moves between them.

Cloud computing is, from the provider’s perspective, about being able to squeeze more value out of hardware and software (using virtualisation, dynamic provisioning and so on) than its customers could by themselves – otherwise there would be no point. Of course, there is nothing to stop any organisation deciding to run its own IT, so there has to be enough reason to hand over the keys the the corporate kingdom to a hosting provider.

For some organisations, this is a simple question of priority. A frequent mantra of business consulting is to stick to one’s knitting – focus on the things the business does well, and outsource those which don’t add value. In reality, though, outsourcing decisions take place more often on the basis of the financials, that is, if third party providers can deliver a service cheaper than running it in house. 

However, cycle for cycle, the cost differential between running systems in-house and using resources from the cloud depends on utilisation. A single server, a cluster or an entire data centre, if used at 100 percent capacity, will be cheaper to buy and run, than to rent. Given that this equation depends on the applications and workloads being run, the server and software architecture, the SLAs and the quality of the operational management, it becomes impossible to compare internal IT with public cloud on costs alone.

So, is it all a simple case of, “Get it right in-house, and use the cloud for the occasional job?” No, not really, due to an additional factor, which could be the make or break for many organisations. Compliance has often been cited as a reason to keep computer systems within the corporate firewall, but this needs to be balanced with the risks of doing so. And all organisations need to take a long, hard look at just how good they are at doing this.

On the visit, what gave most pause for thought was the attention to risk built into the infrastructure. Power supplies from two different utilities, from different routes. Failover generators with enough diesel to last several weeks, and supply contracts already in place. Four different fire sensing systems, pre-action extinguishers and a variety of other mechanisms. Sealed doors and turnstiles, keycards and video cameras, dogs and biometric security that could detect the presence of a pulse (so no more chopping thumbs, Arnie!).

All of which brings back to the question of sticking to knitting. Sure, anyone can run IT systems (I should know). The question is whether an organisation has the time, priority or indeed money to deploy not only servers and software, but all the layers of infrastructure and management required to protect against potential disasters. Or indeed, wants to.


Read more about:

Sign up for our free newsletter