autocratic for the people.

John Siracusa makes an excellent argument for incorporating a second, backup hard drive into home desktop computers, in order to make the inevitable case of drive failure less catastrophic. His argument is Apple-specific, because that's his thing, but it's a pretty good idea across the board. Home computer users are now definitely encouraged to store their entire lives -- music, checkbooks, photo albums, email, porn, tax returns -- in one place, so it's inexcusable to avoid addressing failure of the fragile magnetic drive at the core of it all.

Now, the proposed iMac onboard-backup-drive idea is certainly a great one, and it does fit in well with Apple's position as a purveyor of end-to-end premium experiences (with equally premium pricing). But the backup idea is also dangerous, for the same reason any last line of defense is dangerous: it has to be perfect. In the case of a data backup device, let's define "perfect" as:

1. The saved data has to be up-to-date.
2. The saved data has to be usable.

To the first point, home users shouldn't have to think about backing up their data. It's a hardware issue, it's not their responsibility, and the onboard-backup approach handles this requirement very well if it's set up RAID-1 style (that is, both hard drives are always kept exactly the same -- data saved on one is simultaneously saved on the other). The user need not do anything at all to have an always up-to-date backup, and that's perfect. The requirement is met.

The problem with the onboard-backup approach is revealed with the second, more subtle requirement: the saved data has to be usable. This makes sense; clearly, a backup is useless if the data can't be retrieved, right? But let's rephrase it: the backup drive needs to be even less likely to fail than the primary drive. So we need to ensure that whatever caused the primary drive to fail, it isn't also going to affect the backup drive. This is paramount.

Hard drive failure is most commonly due to mechanical wear of the drive components over time (it's true), which is a problem here because mechanical wear results from usage, and the two drives would be experiencing similar levels of usage if they were mirroring one another. Additionally, we have to assume that in a home environment failure-prediction technologies are not particularly useful because they're too easily ignored -- and besides, many home computers are multi-user so there's no guarantee that every user with data on the drive will know that there's a problem. The kids might not tell mom and dad that the backup is toast until it's too late.

Laying aside the issue of mechanical failure, let's consider other likely reasons for in-home hard drive suicide: electrical problems (lightning or blackouts), physical problems (dropping, hitting or getting the computer wet), or software problems (viruses). For this backup to really be "perfect" and part of a truly premium experience, it must to be able to restore data even in the face of these threats. And, it must be able to survive while not sacrificing any of the requirements we've already enumerated. This is a difficult problem.

Fortunately, I think I not only have a solution, but I have one that fits nicely into the current Apple Airport hardware family. It would be a standalone unit combining a hard drive and wireless router, it would replace the current Airport Extreme, and it would look a lot like the Airport Express. The idea is to combine a home's internet access point (the router) with the backup device, and I believe it would actually solve all of the problems mentioned above.

By putting the backup device into the home's internet access point, the backup device becomes accessible to anyone on the home network. This opens the door to a number of possibilities, most notably that the access point can immediately and continuously warn anyone and everyone connected that there's a problem with the backup. This is sufficient because any machine that backs its data up to the access point must connect to it every so often in order to perform the backup, so anyone affected by the backup's failure -- even people who aren't connected constantly -- are made aware of the problem pretty quickly. Additionally, wary parents can instruct the access point to restrict or lock down internet access if the backup goes south. This would pretty much guarantee an incentive to make the local administrator aware of the problem.

As far as invisibility goes, the backup would occur wirelessly every night (or as a background process during the day), and either way it would go unnoticed by the user. Since the backup data is only transferred between the single computer and the access point, upstream bandwidth is unaffected. Internal network bandwidth is affected to some extent, but chances are the major bottleneck is related to the upstream connection anyway.

This periodic-backup scheme also positively affects reliability -- the backup hard drive is not in constant use as is the primary drive, so its operating life should be significantly longer. In a similar vein, separating the backup drive from the computer decreases the likelihood that accidents involving the power supply or physical machine will affect the backup, and placing the backup in a separate, locked-down device should make it more resilient, in general, to malicious software.

And finally, by making the backup device a separate piece of hardware, consumers need not upgrade their entire computers -- probably inadvertently losing data in the process! -- in order to enjoy the benefits of a reliable backup. Marketing would love it.

This theoretical wireless router / backup device can be made using existing parts. The hard drive would be a low-speed, low-heat, low-power laptop drive -- possibly the 120GB 5400 RPM drive from the current Powerbooks. The rest of the hardware will be straight from the current Airport line.

This device would replace the Airport Extreme, and would retain its price of $199. The marketing would be, "The Airport. Now with secure data backup. Sleep well."

posted by avixe @ 1:08 AM