Opinion: Redundancy trumps resilience for data centre reliability

Shane Chorley of Vox gives us his views on how redundancy is more important than resilience in data centres

August 15, 2011

By Shane Chorley – Executive Head of Vox Telecom’s Core Division

Humans are creating new data at a rate that is hard to comprehend. According to Google chairman Eric Schmidt, every two days we create as much data as the human race did from the dawn of recorded history to 2003 – about 5 exabytes or 5 billion gigabytes.

Five billion gigabytes every two days means a lot storage space – which means data centres around the world are under staggering, unprecedented pressure to cope with the demand.

This is driving some significant changes to the way these data centres are conceptualised, designed and built. Google and Microsoft, for example, are rejecting costly, time-consuming bricks-and-mortar building projects for prefabricated, containerised environments that are quick to build, easy to move – and cheap to replace.

Data centres used to be built for resilience – being able to stay up and running through almost any event. But there’s increasing awareness that this just isn’t possible – there is no such thing as 100% uptime. No matter how hard you try, things will happen to your data centre that are beyond your control. Earthquakes, tsunamis, revolutions, power outages, diesel shortages and financial meltdowns have no respect for your plans.

So nowadays, the emphasis is on redundancy instead. Why put all your eggs in one basket when you can spread the risk across ten?

Two developments have made this possible. First, the price of storage and other hardware has dropped dramatically, to the point where physical servers and storage media are pretty much a commodity.

Second, the technology of virtualisation has come a long way. When Vox Telecom first began running VMware three or four years ago, only a few of our products could be successfully virtualised. Now, we can virtualise almost anything – and there is so much intelligence built into the systems that if anything happens, everything can automatically be rebuilt in a different physical location within minutes.

Depending on your budget, you can now choose to mirror all your servers across two data centres in two different cities, with the guarantee that if anything happens in one centre the other can seamlessly take over.

It’ll be a while before banks move into the cloud like this – it’s appropriate for them to be ultra-cautious – but for many other businesses, especially in the small to medium sector, it’s now effectively a no-brainer. Things like mail servers should all be in the cloud now—if you only have ten users, it makes far more sense to pay a few hundred rands a month for comprehensive mail service than to host your own server.

Moving into the cloud is good for your carbon footprint, too. It’s true that data centres use massive amounts of power, and their absolute power consumption is rising all the time. But thanks to economies of scale – and significant financial incentives – they are able to achieve far greater power efficiency than an environment based on isolated servers.

Things are moving fast in this space, but the direction of change is clear. In five years’ time, it’s a good bet that we’ll all be far more heavily dependent on cloud-based services than we are now.

Tags: , , ,