If you were to believe the hype from the media and certain so called industry experts, you would be forgiven for thinking that cloud computing is the second coming. The answer to all our scaling and usage problems. Everything to everyone.
The notion that the "cloud" is always on, or up, is a myth that is being propogated, irresponsibly, around the corridors of tech. As they say, what goes up must come down and software is no different.
The measure of a redundant system isn't that it stays up 100% of the time, it's how it copes when parts of its systems go down. Can it heal itself? Switch in alternatives? Fail to backup? Reboot parts of itself before disaster is met? All without a disruption in service.
Any long term user of Amazon's EC2 platform knows that instances will disappear without warning every so often. You can be lucky and never experience an outage. Everyone has problems.
Take a look at any providers forums and you'll see examples of people posting about inaccessible instances, unpingable machines, locked up consoles. Amazon, GoGrid, Joyent, Flexiscale to name the top, all suffer the same fate that the more traditional hosting one-machine-one-client have had to deal with for years.
The advantage of the cloud over the historical single server, is the ease and cost at which we, the solution provider, can provision back up plans - cost of hardware is no longer an excuse.
The key to delivering to the cloud platform is realizing that that all your problems are NOT suddenly over. All the usual rules of engagement still apply. The cloud provider is not your miracle worker who can cure all your uptime problems.
As I state in my presentations:
- Always look at the worse case scenario
- Plan for the worse case scenario
- Assume the worse case scenario will happen
- Practice the worse case scenario
Above all, remember this, by the time it comes to complaining to your cloud provider, then you're already losing customers. You are the responsible party, not the cloud provider.
While it is very rare for a cloud provider to go completely offline, it does sadly happen. Flexiscale only recently had a major "glitch" resulting in their whole cloud needing a reboot.
More than likely it will be a network fault, or even a DNS problem that will make a cloud appear completely offline. But either way, assume something will go wrong.
Your mother always advised you against putting all your eggs in one basket, and I bet you didn't guess she was really talking about your cloud provider? the woman was wiser than you maybe give her credit for
The techniques, tools and expertise exist today that lets you utilise more than one cloud provider. Spend a little time learning them (or come to the Cloud Bootcamp and i'll show you) and get yourself a good nights sleep.
Redundancy has never been cheaper or more accessible - that is but one of the true deliverables of cloud computing.