When HN users recommended it
1 HN user recommended this in 2023, in one comment.
20072026
Each tall tick is one HN user's first recommendation. A short tick is the same user again. Click a tick to jump to the comment.
1 HN user recommended this in 2023, in one comment.
Each tall tick is one HN user's first recommendation. A short tick is the same user again. Click a tick to jump to the comment.
> but I've thought about RAIDing some USB thumb drives together
I had a netbook with an irritatingly slow (and a bit small) eMMC drive for its main storage. I had a pair of microSD cards lying around that were faster for many IO patterns, so I bought a third and strung the three together in mini-USB readers (like https://www.amazon.co.uk/dp/B001JPPQS6/) and run Linux off that using its 3-drive RAID10 support. Leaving a bit of space on each for a smaller RAID1 array for /boot (booting direct form RAID10 didn't work), IIRC the 3x8Gb SD cards give a small touch under 12GiB, larger than the built-in 8Gb, and it was significantly faster (more so than I expected). The slow eMMC got used as backup storage, syncing copies of bits from /home and config from elsewhere, so I could rebuild easily should two cards die at the same time.
The main problem with this arrangement was that is consumed all three of the USB ports, if I wanted to access another drive locally rather than over wireless I'd have to shut down and reboot with at least one of the drives hung off a small USB hub. For some reason it didn't work with one of the drives in the built-in SD reader (the arrays would always start in a degraded state on boot) which I never got around to looking into working around (perhaps a kernel module was missing from initrd that was needed to be able to access the hardware?).
I experimented with RAID5 and RAID0 also. 5 gave a bit more space but was a lot slower for write operations, at least in artificial benchmarks, I assume due to read-before-write issues and the machine having little RAM spare for cache that might mitigate this. 0 did not give notable read performance benefits (though obviously gave much more space and measurably faster writes), perhaps all the ports running of a single internal USB2 hub was a limiting factor.