Bonus points if this causes my phone to make Airbus ECAM noises.

Sure, my desktop browser can emulate a mobile device, but if I have a real live mobile device right in front of me, why not use it?

Of course, all of the usual messages about data management and backups apply here. RAID is not a backup strategy. You should have a plan for point-in-time restore, and you should have a plan for your _entire_ RAID array being wiped out, including off-site storage.

But, for hot storage, large drives create risks that are almost certainly not worth the benefit of needing fewer drives overall.

Again, if you just want a big place to stash some data that you genuinely don't care about losing—a "convenience copy", if you will—then maybe a large high-capacity drive is worth it. Such a drive may also be worth it as part of a long-term archiving plan.

But what if you actually have tens or hundreds of terabytes of data? You're going to need more drive bays in your NAS, period.

In short, you want to keep rebuild times down. You want to get the array out of a degraded state as quickly as possible, and that means that waiting to copy 12, 16, or even 24 TB on to a new drive can be an unacceptable risk.

If you're running raidz3, for example, you may be able to withstand this. Otherwise, you may find yourself testing your backup plans.

This effect is more acute when you have a smaller number of large drives, rather than a larger number of small drives.

Rebuilding a RAID array necessarily stresses the remaining drives. Sometimes, this leads to cascading failures, where the stress of a rebuild takes out another drive before the rebuild has completed.

The trouble is, as drive capacities increase, so do rebuild times. And that's assuming you even have a replacement drive on hand (or even better, a hot spare in the array). In the worst case, the array runs in a degraded state for a while, before eventually beginning the rebuild.

But here's the thing: while hard drive reliability has greatly improved since the days of 500 MB drives and stiction and IBM Deathstars, losing a modern high-capacity drive is still a possibility and would be enormously impactful.

"Don't worry, I'll put it in a RAID array!", you might say.

Okay maybe we need to include frequency on the map as well.

lbfoster.com/rail/frictio...

I know there are other benefits to PSDs, like improved station climate control, but the aesthetic impact is a very real thing!

The DLR doesn’t even have any form of guideway intrusion detection, unlike SkyTrain!

And nowadays, if you’re really concerned about intrusion detection, it’s easy enough to fit trains with forward-facing LIDAR.