Increasingly businesses are being told to put their data in the hands of a cloud provider, either in the form of a backup to on-premise system or a full deployment.
With more and more businesses moving to vendors like Amazon Web Services (AWS), Google Cloud Platform and Microsoft Private Cloud, we're increasingly seeing the vital infrastructure of multi-billion dollar corporations placed in the hands of another multi-billion dollar corporation.
In some ways it makes sense. If you make cars, you're not a computer expert - so why manage your own IT structure? But on the other hand - what if something goes wrong?
Last week we saw what happens. AWS fell over on 28th February and it was all down to a simple typo in a piece of code - in the billing system of all things. The outage took out not only thousands of websites, including some big names (though oddly, not Amazon itself), but also the AWS Service Health Dashboard (SHD) which is supposed to be there to tell you there's a problem in the first place.
The whole thing was resolved relatively quickly but it did require for Amazon to "switch it off and on again" at their end which is no small feat for a huge multi-national network of cloud servers.
OK. First of all, let's be clear, it's extremely unlikely. To call yourself an enterprise cloud vendor, you're going to need fail safes on fail safes to ensure that you don't let customers down. In 2015 for example, Amazon boasted a 99.9974 percent uptime - and considering that's the big daddy, it's pretty impressive. In 2014, Google's storage was down for a total of 14 minutes. That's it.
In fact, that a big cloud outage like this happened at all is so improbable, it required me to go back and rewrite large chunks of this article that was already in progress. But generally, the cloud doesn't fail, or at least, not very often, and even then only for short periods of time.
Bu when outages do occur they are business critical. Remember last year when JetBlue's systems, provided by Verizon, went down causing flight delays that went on for over a day? Or when the (very private indeed) cloud of the Link network caused cash machines to stop working?
As a consumer, it's annoying. As a customer though, it's a disaster both commercially and in PR terms it can take years to recover.
And it's that word "disaster" that's the crux. Too many companies don't take responsibility for their disaster recovery. Whether that's alternative systems to switch to for business continuation, or a "plan B" (which may, shock horror, involve pen and paper), the need to not live and die by your cloud is a painful reality.
So do you know what you'd do if your cloud evaporated?
First of all, don't panic. the chances of an extended outage are almost zero. The closest we've come to that is the Denial of Service (DDoS) attacks that went straight for the DNS servers instead of attacking the end targets. That was both unprecedented and unprepared for.
But now it has happened, it's just that little bit less likely to happen again. Being wise in hindsight is a frustrating thing, but it's sometimes what has to happen. Just as happened with Heartbleed and encryption, it was an example of a weak link in the chain that has now been plugged. Hopefully.
But do you have business critical information duplicated elsewhere?
Take for example, a supermarket. These days, all the big chain stores have internet enabled tills that check prices, control stock and predict demand, all off-premise, in the cloud. And have you ever been there when that system fails?
Headless. Chickens. The days when the prices were on the items on little sticky labels is long gone, and when the computers fail, no matter how many times you see it, it feels like nobody has learned contingency to keep trading.
So this is the battle cry. Be prepared. Be prepared for the worst that hopefully will never happen. A simple on premise solution that provides contingency to your network could be the difference between closing down, and riding the outage like a champion.
This sort of hybrid set up can also make your business more agile, with low latency on-premise response for the big stuff, and slightly slower (though these days almost imperceptible) cloud calls for the less common, or less urgent stuff.
This shouldn't put you off the cloud. It offers so many benefits that an on-premise solution just can't compete with. All cloud providers have strict service level agreements - Microsoft Azure offers 99.95 percent on Virtual Machines alone.
The point is, nothing is infallible and working with your cloud provider to ensure business continuity should be your responsibility. You can guarantee they'll have some great ideas and will be more than glad to help. But when something goes wrong and your customers see a 404 screen, it's already too late.
Including a 15-inch Intel Core-powered device weighing less than a bag of sugar
Tuomo Suntola's ALD technology extended Moore's Law, but was only adopted by chip-makers in 2007
Trump proposes a $1.3bn fine and a round of firings to un-bork ZTE
Findings could mean new optical frequencies to transmit more data along optical cables