Original URL: http://www.theregister.co.uk/2009/10/18/danger_failure/
Danger lurks in the clouds
It's not just a backup, you know
Comment The failure of Microsoft to safeguard data synchronised from Danger's Sidekick devices on T-Mobile's network has thrown up important questions about cloud-based storage, along with insufferable smugness from iPhone owners.
Most cloud-based services aimed at consumers are still on the backup side of things, offering to hold a copy of your data - the original of which is held on your own computer or elsewhere. But with cloud services increasingly wanting to process your data too, the failure of Danger's service could well be a nasty omen of things to come.
Quite what went wrong at Danger we still don't know, though AppleInsider has come down firmly on the side of deliberate sabotage quoting "sources" that confirm:
"Someone with access to the servers at the datacenter must have inserted a time bomb to wipe out not just all of the data, but also all of the backup tapes, and finally, I suspect, reformatting the server hard drives so that the service itself could not be restarted with a simple reboot."
AppleInsider also reaches new levels of self-satisfaction when explaining how such a thing could never happen to an iPhone as backups of data are held on an obligatory PC rather than synchronised with the cloud, in contrast with Microsoft's MyPhone service which makes a local backup optional.
Having a local backup might seem an obvious point, but anyone who's worked with synchronisation software will know that having multiple copies of the same data is a lot less useful if those copies are synchronised together - errors and corruption get synced just like everything else. On this occasion local copies might have helped, but next time that might not be the case.
The issue here was that the Sidekick is designed to default to the server when something goes wrong. If the device isn't shut down properly, or encounters any corruption of data, it just gives up, cleans down and replaces everything with the data set stored on the server - only the server isn't there right now.
If nothing goes wrong then the Sidekick should operate without any problems, which is why T-Mobile is advising users: "Please DO NOT remove your battery, reset your Sidekick, or allow it to lose power."
So the problem isn't just the catastrophic failure of the data centre, it's also the fact that Sidekick devices are not conversant with the concept of a server failure. MobileMe, MyPhone and Ovi could all disappear in a flurry of commercial short-termism and no one would be as inconvenienced as those poor Sidekick users.
But as connectivity becomes more ubiquitous, applications will increasingly expect it to be there - anyone who's tried to navigate to somewhere without mobile coverage using Google Mobile Maps will know the frustration as the maps disappear even if previously visible. Mobile applications increasingly expect servers to be there, and have little idea how to cope without them, even though mobile coverage often fails to live up to the promise.
Sidekick aside, today's mobile systems are designed to be stand-alone - the unreliable nature of wireless coverage demands they be able to operate with limited to no connectivity, though as reliability improves that could well change.
T-Mobile is now saying that users shouldn't give up hope: the operator will be in touch over the next two weeks to restore the lost data, or send $100 in T-Mobile credit if the data has gone forever, so some backups have obviously been located.
Cloud-based servers are still more reliable than most of the kit knocking around users' homes - the life expectancy of an Apple Time Capsule, for example, is just over 17 months according to the Time Capsule Memorial Register, so even those who are backing up locally shouldn't be too smug.
Microsoft's failure to manage customers' data is embarrassing, but will probably do more damage to the concept of cloud-based storage than it should. Most users should be more concerned about corrupted data than outright failure - the latter being easier to spot, and fix.
Testing an early implementation of SyncML your correspondent had the gruesome experience of watching his entire address book disappear, entry by entry, as a corrupted server echoed its pain. On that occasion the only recourse was an elderly printout and a lot of typing, demonstrating that at the end of the day there's still only one medium that can really be trusted. ®