The recent outage of Amazon’s web services highlights issues that have yet to be resolved in the cloud. I’ve been reading a lot by people spouting off about how the companies that were affected should have planned for problems, along with a number of smug online services/sites that survived retelling how they’re so smart. Which is fine, good for them, but the idea that seems to be that your service/site dying wasn’t Amazon’s fault. Well, blatantly, it was.
It may be OK for a large, highly technical company to spend a lot of time, money and brain cells to make sure their data is redundant, but that’s not going to work for most companies or consumers. The point of the cloud is to be able to outsource your software/storage/computing needs and pay someone to worry about that for you. All you need to worry about is being able to connect to the internet.
What we’ve all been led to believe is that even if a nuclear warhead dropped on one of the data centres, cloud operators have many, so you just switch to providing the file from another state or another country, the end user may see a slight blip and the files may take a few milliseconds longer to reach them, but that’s it. So how did a failure on this scale happen? No one really knows (outside of Amazon).
We can all buy a computer and store our files on it, it’s the ultimate distributed system, practically impossible to knock out more than a handful of nodes at any one time, not worth hackers trying to attack all but a few, easy. The cloud is supposed to take away the need to buy and maintain hardware and offer it at a better price, because I don’t need a whole machine to myself most of the time. So they can buy lots of machines and get bulk discount, then share the machines out so individuals only use the resources they need. What should be happening is that rather than running on one server, my files and apps should be spread thinly across many and, should one die, the others take up the slack and then find a replacement (much like how RAID 5 works). Tape backup (like Google had to use after a recent Gmail outage) shouldn’t be necessary as there are redundant copies on several machines in several sites.
That won’t help if the data centre goes down though (and they do, despite their claims of backup power, multiple lines of power and comms, there’s always something they haven’t thought of), but that’s OK, because all the big boys say they have multiple, so they can mirror this process over multiple data centres, even around the world.
That’s the vision we’re being sold, but blatantly it isn’t happening. Somewhere, the distributed computing has failed. What the recent outage shows is that Amazon and the rest aren’t doing this and they should be. Unless a large asteroid wipes out Earth, my data and apps should be fine, that is what all of the cloud services seem to be offering and should be achieving, until then, they’re just another liability, all you’re doing is moving the problem from in-house to somewhere you have less control and which you can only access via a helpdesk.
That’s not the only problem the cloud has highlighted in the last week or so, there has been more news of security concerns at one of the most high-profile file hosting sites, Dropbox. They recently changed their Terms of Service to state they will hand over your files to the US government if it asks, not so much of an issue, maybe, but it highlighted the fact that employees obviously do have access to unencrypted versions of the files, which went against a statement on their FAQ page. Their software was also found to have a security flaw that could allow anyone full access to all of your files. The concerns have spread so far in the online community Dropbox have responded with a blog entry addressing the various concerns.
I don’t think Dropbox are unique in this, or even any worse than any other online file hosting/sharing/backup company. I’ve said before I don’t particularly like the idea of sharing my files on these sites, so I encrypt the ones I’m worried about before I load them (so even if the employees, authorities or anyone else get access to them they’ll be useless anyway).
Add these to the fact that most cloud services make it hard to get all your data out, which makes it hard, if not impossible, to move (thereby actually giving you less choice as you’ll never be able to move) and the growing size of the files we’re working with (photos, videos, games) and doing things remotely becomes even less desirable as the networks are unable to sustain it (backing up even a modest song collection could take days). What we’re left with a situation where we could do with being able to produce our own cloud, at least for consumers. It’s something I’m going to take a look at.
The question is, with the rise of the smartphone, the ever growing portability of processing power and storage is quickly increasing. Why keep all your files online so you can have them wherever you are when you can keep them in your pocket, wherever you are?