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

It does change due to deletes. Otherwise, it would be obscenely expensive to never delete data. Garbage collection is hard but since it’s async it’s also possible to do efficiently while on SMR.



At this scale, does it really matter? Right now everything I record in the cloud has an infinite history attached to it, so nothing is ever really "deleted" as I understand it, the index merely points to the latest version of a file.

If the deletion is meant as a security feature, then you can "delete" a file by making it encrypted and deleting it's decryption key.


Most clouds these days do have a limit on maximum version retention. Otherwise you'd effectively get infinite storage for free.

If a cloud storage provider doesn't charge for infinite version, that's not very different from offering "unlimited storage" plans: They have pretty good statistics of how much the average user really uses on an "unlimited" plan and will price their services accordingly. Revision history just adds another term to the same equation.


Agreed with this point! At this scale more aggressive garbage collection doesn't save you much actual money.

In the early days at one such provider I know that garbage collecting deleted user files was considered a relatively low priority cost-wise.


The more data you have, the more savings. Savings at 1 exabyte are way different than savings at 10, 20, 50 exabytes. This ends up being a lot of storage racks. File revisions is a minimal part of the story.




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

Search: