RAIDING PARTY
Setting up SSDs properly used to be a pain, both on Linux and Windows. Filesystems and partitions have to be aligned to 512 (or sometimes 4096) byte sectors, otherwise the drive will not perform well. Nowadays, partitioning tools and filesystems will do all of this for you. Run
There are other RAID levels, too. A RAID6 configuration might have four drives that use maths to spread parity data such that any two drives may expire. When a drive fails it should be replaced immediately. You can mount an array in degraded mode (so long as too many drives haven’t failed), but you won’t have any redundancy until the array is reconstructed. So cautious data fiends will keep spare ‘hot’ drives enrolled in their arrays that can be summoned into service. Rebuilding can take a lot of time (potentially days when terabytes of data are involved) and deeper RAID levels involves a lot of disk hammering, so there’s potential for cascading failures.
$ sudo fstab -l
Many motherboards have their own, proprietary RAID implementations (pejoratively termed FakeRAID), and hardware RAID controllers are still used in enterprise. However, these aren’t strictly required now: the parity calculations involved are easy work for modern CPUs, and software RAID is entirely viable. The old (and still perfectly good) way to do RAID on Linux is through mdadm, which presents the array as a single device that you can format however you like.
if you want to be sure. In order to maintain the sort of transfer rates you see in the tech tabloids, you need to make use of the Trim feature. This is where the filesystem tells the drive which storage blocks are no longer in use, so that it doesn’t need to worry about them any more.