ZFS, Actually Storage you can finally reason about.

Mirrors

concept
depth 2/3 − deeper + surface

A mirror is the simplest disk team: two (or more) disks holding identical copies of everything. The space math makes it look like the wasteful option—you "lose" half your disks to copies. Actually, the space you give up buys the three things the clever alternative can't sell you: reads answered by whichever disk is free (so two disks are roughly twice as fast to read from), the quickest possible recovery when a disk dies (just copy from the healthy twin), and the most freedom in all of ZFS to grow and rearrange later.

Where mirrors are simply the right answer: pools whose workload is random I/O—VMs, databases, container volumes. The pool stripes across vdevs, so six disks as three mirror vdevs gives roughly 3× write IOPS and 6× read IOPS of a single disk; the same six as one RAIDZ2 gives roughly 1× (see RAIDZ for why). This is the whole "mirrors for VMs" folk wisdom, derived.

Operations worth knowing exist: zpool attach converts a lone disk into a mirror (the retroactive fix for a single-disk-vdev regret) or adds a third member to a pair—and resilvering a mirror is reading one healthy disk, not reconstructing from the whole vdev. Growth path: replace both members with bigger disks one resilver at a time, or add another mirror vdev to the pool.