Native ZFS VDEV for Object Storage (OpenZFS Summit) (zettalane.com)

That’s brilliant! Always amazed at how zfs keeps morphing and stays relevant!

How does this relate to the work presented a few years ago by the ZFS devs using S3 as object storage? https://youtu.be/opW9KhjOQ3Q?si=CgrYi0P4q9gz-2Mq

I do not get it.

Why would I use zfs for this? Isn't the power of zfs that it's a filesystem with checksum and stuff like encryption?

Why would I use it for s3?

FS metrics without random IO benchmark are near meaningless, sequential read is best case for basically every file system and it's essentially "how fast you can get things from S3" in this case

Could someone possibly compare this to https://www.zerofs.net/nbd-devices ("zpool create mypool /dev/nbd0 /dev/nbd1 /dev/nbd2")

How suitable would this be as a zfs send target to back up your local zfs datasets to object storage?

I am curious about the inverse, using the dataset layer, to implement some higher level things like objects for an S3 compatible storage or pages directly for an RDBMS. I seem to remember hearing rumblings about that but it is hard to dredge up.

Exciting stuff, but will this be merged? I remember another similar effort that went nowhere because the company decided to not proceed with it

It doesn't look like the source has been released, nor any documentation outside this blog post and presentation. Is there a plan to open this up past what is used by MayaNAS and Zettalane's cloud offerings?