SSDFS Is The Newest Linux Filesystem & Catering To NVMe ZNS SSDs

[phoronix]

Phoronix: SSDFS Is The Newest Linux-Filesystem & Catering To NVMe ZNS SSDs

Sent out for review on Friday evening were 76 patches implementing SSDFS, the newest open-source Linux file-system and catering to flash-friendly drives and particularly those with NVMe Zoned Namespaces (ZNS) support...

https://www.phoronix.com/news/Linux-SSDFS-NVMe-ZNS-SSDs

[LinAdmin]

That sounds pretty good. However, where can one find the sources of the ssdfs.tools?

[Mitch]

This is a bit random, but I'd love for a filesystem to let me configure parity writes. I'd be happy to have checksums and spare 5% - 30% space and write performance just to add parity, so the filesystem could detect and heal bad data on a single disk.

[erniv2]

That sounds pretty good. However, where can one find the sources of the ssdfs.tools?

GitHub - dubeyko/ssdfs-tools: SSDFS file system utilities was in the opening post.

[Velocity]

How does it compare to Copy-on-Write filesystems such as ZFS?

[erniv2]

This is a bit random, but I'd love for a filesystem to let me configure parity writes. I'd be happy to have checksums and spare 5% - 30% space and write performance just to add parity, so the filesystem could detect and heal bad data on a single disk.

That is not needed since ages HDD´s carry ECC bits and spare space for cluster failures, once you encounter a fatal hardware failure the data is lost no matter what, if you have a journal parity or whatever software solution, in fact trying to rebuild a fs on a faulty drive will destory your intact data for sure.

In that case only the multidisk raid solution offers you safety to swap the faulty hardware.

[patrick1946]

How does it compare to Copy-on-Write filesystems such as ZFS?

I would imagine it is copy on write too because to my understanding on ssds you don't write on the same page but write on a different page.

[EphemeralEft]

Not another file system. Especially if it’s CoW. It seems like an NIH pandemic.

Also, having better performance than existing solutions while lacking many of their features is not noteworthy. Those features come at a cost and if you don’t have them you’re just doing what other solutions are doing but worse.

[skeevy420]

Same Shit, Different File System


But, for real, this sounds...reads...pretty neat.

[Velocity]

I would imagine it is copy on write too because to my understanding on ssds you don't write on the same page but write on a different page.

Yes, so does this provide any benefit over the tried and tested ZFS filesystem which is the most advanced single system filesystem on the planet?

As i understand, copy on write means that filesystems will not overwrite the same block when appending a file or modifying a file, but would use free LBA space. This enables the SSD firmware with its Flash Translation Layer (FTL) an opportunity to keep things tidy and not pollute the FTL with many 'overwritten' blocks which are problematic in terms of garbage collection, free NAND erase blocks and the resulting write amplification. CoW filesystems therefore are very much suited to SSDs employing such an FTL.

Is there any benefit this filesystem provides over the already existing ZFS filesystem? Or is this just a GPL abomination of a superior filesystem, and it just comes down to the license wars i've witnessed for decades now. At least ExFat works on multiple operating systems so it is cross-platform. This will be linux-only, not even supporting BSD?