Every time someone bitched at me for not having a "cloud-based strategy", I kept asking how many 9s of reliability they thought the cloud would deliver.
We're down to 3 nines so far. A few more hours to 2 nines.
"The cloud", as I understand it, is the ubiquitous, cheap and near-instantaneous availability of computing power; as in minutes instead of hours or days for new servers.
"The cloud" is not (and never has been) a cure-all for reliability issues. It's just as easy to have single points of failure as any other hosting strategy, and is just as easy (or difficult) to plan for. Companies that have planned for high availability with multi-region or multi-provider strategies will continue to be available, regardless of whether or not they are using "the cloud".
That's an issue with one service in one region offered by Amazon Web Services, not "the cloud" as a concept.
Use this as an example of the reliability of EBS (or if you want to broaden the scope, Amazon Web Services) all you want, but this says nothing about "the cloud" as a concept.
If your business is amongst the chosen few that can justify the cost to guarantee any number of nines then your availability strategy involves multiple vendors anyways.
The cloud is not for all businesses.
Whether Amazon can be part of an availability strategy has nothing to do with the number of nines.
We have our business website hosted out of the Amazon cloud. Our primary servers are actually located in their affected data center. But we also have a great data team behind it, so we aren't being (to the outside observer) affected at all by the outage.
Cloud is vulnerable? Of course it is. So plan accordingly.
The same steps you would take in your own datacenters to ensure high availability would work in the cloud to ensure the same availability so I'm not sure what your point is. Measuring the availability of a few zones from one provider and broadly labeling the cloud as unreliable is a flawed argument. Netflix, for example, is entirely on AWS and is still running well today.
> The same steps you would take in your own datacenters to ensure high availability would work in the cloud to ensure the same availability
If I'm engineering the same steps in the cloud as I am in the data center, then I'm going to skip a step and just engineer the data center, because adding machines on demand is not rocket science. But maybe that's just me.
Do you already have a proven API and hardware provider who will provide you with said machines? Got them racked up and powered? If not, there's something missing.
Hmm, "bitched at you" has the ring of feeling persecuted because you didn't jump up all dreamy eyed at the latest buzz word trend. Occupational hazard I suspect.
If someone says to you "We need to improve the efficiency of our IT by adopting a cloud based strategy." Rather than ask them the 'meta' question of what sort of reliability guarantees they have, have an actual and honest talk about what IT costs and why. And perhaps they will relax their uptime requirement which will let you reduce your costs, or they will come to understand what the costs are for the level of uptime you're providing. Annual reviews of those questions (how much downtime can we tolerate, how much are we paying to achieve our current availability?) should be de rigueur.
"The cloud is not for all businesses."
Of course it isn't. However it can (and does) run some businesses more efficiently. And while Quora might be down for a day while folks at Amazon scramble to fix what ever it is they did that brought it down, their "business" won't change all that much. There will be no mass exodus of users because they could get their questions answered for one day. Now if you take someone's email away for a day, that is real money, or if you take away their ability to connect to the Internet period.
For something like icanhazcheeseburger even two 9s is probably good enough. That would be offline for 3.6 days of the year.
We're down to 3 nines so far. A few more hours to 2 nines.
The cloud is not for all businesses.