Hacker News new | past | comments | ask | show | jobs | submit login

The main questions would be durability and availability. Maybe only because I am scarred by experience but "on-prem" seems to me quite obviously the worst possible way to do it.



Why not both? Store it locally for ease of access, maybe with a solid state cache, backup in the cloud for durability. Storage is cheap nowadays. The more interesting question is availability - at the current HDD sizes, "classic" RAID is not sufficient.


> at the current HDD sizes, "classic" RAID is not sufficient.

Eh, RAID 6 is fine. Your chance of losing two extra drives in a two day rebuild isn't that high if you've been testing your drives regularly.

And if you haven't been testing regularly, even small drives would eventually have some scary risks of dying.

That said, 50% more drives with raidz3 is a lot safer than RAID 6.


It’s not as clear cut as this. Raid6 on how many disks? raid6/zraid2 with 6 disks total might be fine, but 20 disks is not.

One of the more common scenarios with consumer grade hardware is a degraded array getting completely lost while being rebuilt.


> Raid6 on how many disks?

The same number of disks as you were using when the disks were smaller.

I'm comparing like-for-like except disk size.

> One of the more common scenarios with consumer grade hardware is a degraded array getting completely lost while being rebuilt.

Which is why I mentioned regular testing. If your disks are on the verge of death it doesn't matter how much parity you have or how big they are, a rebuild has too much chance of failure. You need to make sure they can handle a moderately heavy load.

But if you're putting them under that load regularly, a rebuild is very unlikely to be the last straw for multiple drives.




Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: