Hacker News Reader

Comment by iberator | Hacker News Reader

Which manfucaturer is the best and worst one?

It almost doesn't matter. There's three choices for spinning drives and all storage devices fail (even SSDs), and you need a plan for that. There's variance in failure rates and warranty periods, but most (new) drives last the warranty period, and used drives is more complex than just manufacturer --- how were they used, stored, and handled before resale probably makes more difference than who made them.

IMHO, it's more important to consider correlated failures rather than worry about getting the best or avoiding the worst drives. Try to avoid running an array that's built from same model, same firmware, same build time, same power on hours, same workload. Every so often, you get things like drive disappears when power on time overflows, or a manufacturing error that makes most drives fail after N weeks of use, having all of your drives in the same part of their lifecycle makes you more likely to experience a catastrophic failure that failure rate analysis wouldn't pick up.

Picking SSDs carefully bears more fruit, there are many more makers and wider variance in performance and reliability as well as characteristics during failure: everyone says SSDs go read only during failure, but my experience has been that lots of SSDs disappear from their interfaces during failure; you might reasonably have less redundancy if you have confidence the SSD will remain readable for recovery if it fails.

You have to be careful here; mismatched drives can tank array performance.

It's all tradeoffs. For my personal use, array performance would have to be terrible for me to notice, but durability is important. I've run systems at work where durability isn't required, but performance is.

If you run a fully matched array, you may well get better performance, but you also may wake up one day to all your disks disappeared. There's other ways for all your disks to disappear too, but same firmware, same firmware bug feels more preventable than power supply failed very hot.

I mean in basic terms like 5400 vs 7200 rpm spinning rust

https://www.backblaze.com/blog/backblaze-drive-stats-for-q3-...

> For those of you tracking the stats closely, you’ll notice that the Seagate ST8000NM000A (8TB) is a frequent flier on this list. The last time it had a failure was in Q3 2024—and it was just a single failure for the whole quarter!

Do not buy Kioxia. If your drives dies after a few years, but can come back to life with an firmware update. They will not give it to you unless you have a support contract.

Solidigm have all their firmware available for everyone their website.

I just get whatever is cheapest with a warranty of one or more years. RMA ability is most important when buying these kinds of heavily used decommissioned drives. ZFS RAID Z2 is absolutely key to preventing data loss here.

Check the Backblaze quarterly reports.

Reply on news.ycombinator.com