In the last two weeks of June two power interruptions in one of Amazon’s US data centres resulted in outages for some of its AWS customers. The internet and in particular Twitter were awash with people passing comment or, in the case of some of those affected, bemoaning the impact. Post mortems followed – and discussions on the web continue. Some customers moved their services back onto their own, or co-located, floor plates and a number of arm chair pundits made statements that it is always better to own hardware than to rely on cloud providers.
Having run data centres for many years and suffered my fair share of hardware and power failures I am somewhat more laissez faire in my response. My view is if you run data centres, whoever you are, you will suffer failures. Call it law of nature or physics it doesn’t matter which – it is the nature of the beast. So, that said, what is the key take away? Well, it is the same for Cloud based services as it is for any services located in data centres on your own premises. You have to have robust Business Continuity Plans (BCP) in place – your business processes have to be able to continue in the absence of your IT systems. You need to understand the criticality of your applications and if your IT systems are truly mission critical and your business process cannot continue without them, then you must architect your systems accordingly, irrespective of where they are hosted.
Cloud service providers may be less prone to service interruptions than data centres run on the cheap by companies trying to scrimp and save on IT costs, but they aren’t immune and shouldn’t assumed to be invincible. Cloud Data Centres are indeed fallible too.