Risks of running in the cloud
I think we’re beginning to see cloud computing mature a little bit. There is a long ways to go, but I am detecting more sober thinking on a wider scale. The events of last week, where many people were affected by Amazon’s outages, are helping clues sink in a little bit.
There are a lot of risks that don’t go away when you use cloud platforms. There are a lot of new risks that you’ll encounter. And yes, some problems are outsourced. But clear-headed thinking on “what is cloud computing about and what is it good for?” is still largely nonexistent on a broad scale. The din of the hype machine has most IT folks’s heads ringing and they can’t think straight.
Is that an over-generalization? I don’t think so. Go look at Twitter and I think you’ll see a pretty good representation of the general state of people’s thought processes around cloud computing. You can look at the four-page special advertising spreads and pull-out sections in the Wall Street Journal for the source of the confusion. A recent pull-out section told me that cloud computing is going to revolutionize even the nursery — who wants a baby monitor when you could have a laptop sitting in your nursery with a webcam watching the babies? That was over the top, but not too far. Most of what’s written about “cloud computing” is both ridiculously hyped and completely off-topic; writers will slap “cloud” on anything tech-related to get eyeballs on it. And that’s causing a general muddiness of the waters.
But this week, I heard some people thinking about questions like “if I rely on an external services provider for something, and there is an emergency that affects a lot of their customers, will I have to wait in line to get help for my own systems?” That’s sober thinking. We need more of that. A monoculture of any type is a problem. If I were building something on a cloud platform, I’d ensure that it was diversified not just across availability zones, but across completely different providers.
This is one reason why we need standardized APIs for managing cloud resources, but that’s a digression.
I know you’re talking here about IT-savvy consumers of cloud services here, perhaps up to the C-level executive. And yes, it’s good that such customers are now seriously thinking about backup plans, temporary alternate sites, graceful degradation, and all the painful little things they don’t want to plan for when their magical, always-on cloud provider turns not so on for a while.
But I don’t think the average person has a terribly clear notion of what “the cloud” even means. Have you seen those couple of recent Microsoft commercials? Obviously Microsoft is late to yet another party, and they want to show everyone they “get it” and can do cloud services as well as anyone. So what two applications do they show, in each case with a catch phrase, “To the cloud!”? One has a mom fixing up her unruly family’s group portrait by copying-and-pasting together two shots in a photo editor, and the other is a bored couple at the airport using a remote desktop app to watch a TV show recorded on their PC at home. Neither one of these is a cloud service, dang it! I guess it’s understandable, because when you get down to it, the cloud is pretty boring, and it doesn’t really make good television. “Look! I can work on a spreadsheet from two computers without putting it on a flash drive or emailing it to myself!” is not really a riveting plotline for a 30-second spot.
Jeff Saxe
25 Apr 11 at 1:08 am
Yeah, “the cloud” is just absurd. It meant something before the marketing folks jumped on it and started slapping the label on anything with an on/off button. I think “the cloud” will burn off and drift away, and what will be left is the realization that controlling infrastructure through APIs is a good idea.
Xaprb
25 Apr 11 at 2:46 pm
You have it right. “The Cloud” could be a useful tool, but really it shifts the risks and costs rather than avoiding them. How much down time will it take to erase the “savings” many of the companies have gained by not managing their own infrastructure?
If people are going to use it for critical systems they have to accept that they no longer have control over the physical infrastructure so they have to do a better job of designing their software to handle the possibility that the “hardware” will go away. Or slow down due to over-subscription (which is a key piece of the cloud that we haven’t run into problems with yet as far as I know but I guarantee that is coming. Cable was a blazingly fast internet access method for a while). That takes more highly skilled developers, which means money.
Once again, something new came out that made creating software systems appear easy. And the crowd went wild for it. And once again, it’s limitations demonstrated that writing GOOD software systems is not easy. No matter what tool you use.
John
25 Apr 11 at 4:49 pm
I’ve been beating the over-subscription-in-the-cloud drum for a while, but I predict that when it becomes a reality, other people will get credit for being the sooth sayers :)
Xaprb
25 Apr 11 at 6:50 pm