-
Notifications
You must be signed in to change notification settings - Fork 707
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Revisit API of storages #2674
Comments
The
I am still not sure about this, we still need a way to offload things from memory, otherwise, a long-running scape would fail at some point. |
As long as the memory-only thing is opt-in and the default is filesystem-backed, I'd argue we don't 🤷 |
I see, so the default would be the file system one. Then I am not sure what benefits this would bring, pretty much all the problems we had were about the file system part. And that implementation would still require a lot of in-memory caching most likely, I would even say it would end up as the same as the current memory storage. |
I'd personally be opposed to making fs-backed default, especially for requests (as we've seen the amount of troubles that causes, especially during concurrent accesses)... But we can definitely optimize some parts further I think
Should work with current memory-storage, does it not? memory-storage definitely supports it to my knowledge (.s.getOrCreate() -> new UUID) |
This would work out of box already if we remove the storage manager cache, its not about the storage implementations, its about the fact that you calling |
There are multiple considerations:
Dataset.open
)The text was updated successfully, but these errors were encountered: