> ZFS is great too.

Could you give some of your personal reasons for saying so? So much online sources just say how great it is without any specific detail. It's not like XFS on Linux has some obvious problems - it is rock solid too.

Let's say I want to run several different operating systems simultaneously from a single big ass SSD. With Linux, XFS and Xen, I can try to set up LVM and coerce Xen to use those logical volumes for VMs, but it is pain to get there with the current state of XFS and Xen documentation. Is this possible/easier with FreeBSD and ZFS?

Just sharing my personal experience.

Context: I use XFS as the default file system on all HPC compute nodes, and few other servers that run CentOS7. I built a FreeBSD+ZFS backup server recently to mirror our primary NAS.

The experience with ZFS was so smooth and logical. All I needed was the ZFS section from the FreeBSD handbook, and I actually understood what was going on underneath to a reasonable extent. No magic - "run this command" without explanation. Once spools are set, ZFS datasets provide the equivalent of logical volumes you were asking for. You get a lot of tunables at that level, and they come with sensible defaults. The whole process is simple, heck you can even make a file act as a zpool for learning or testing purposes.

Once done, I was testing performance rudimentarily and saw that it was almost saturating the network pipes beating my expectations without any time spent on custom perf tuning.

And I get snapshots and online compression, SW RAID out of the box.

I may be able to achieve most of it with XFS, some specific hardware and lot of hours of tuning, but it will never match the experience with ZFS.

ZFS blog post: https://aravindh.net/post/zfs_fileserver/

Performance: https://aravindh.net/post/zfs_performance/

Thank you very much for this. I'm thinking about using ZFS on a single big ssd, as a storage for several simultaneously running VMs. I'm aiming for easier administration and advanced features like error detection/correction on the drive. Would ZFS make sense in that case (single physical drive)? Or do I have to put in more physical drives to get the actual benefits of ZFS, like you did with NAS backup?

> Could you give some of your personal reasons for saying so?

For one example, ZFS checksums spot all manner of erratic, intermittent idiocies--memory corruption, data bus corruption, disk controller corruption, etc.

No other filesystem even tries.

Would that work also if I had only a single big drive? Say, divide the space on the drive into 2 parts and let ZFS use those parts as RAID1 disks?

AIX sort of died in the 2000s. Back in the late 90s the only reason to have AIX systems were as gateways to the AS400 systems. You didn't generally use AIX systems for anything else back then.

I don't do Aix since 2003, does it still follow Windows model in how .so are created with export definition files?

What you mean by that? I stay away from Windows ;-)

Back in those days, Aix did not use the ELF binary model, rather IBM went their own path and imported the Windows concept of definition files and import libraries to generate shared objects.

Aix shared objects had more in common with Windows world than with other UNIX variants.

I've started with 5.1 therefore I can't relate to that.

Thanks for the answer, I got curious and decided to dive again into Aix docs, and apparently export files are still a thing.

https://www.ibm.com/support/knowledgecenter/en/ssw_aix_72/co...