Makes no sense. First ZFS needs direct access to the disk without any abstraction layer in between for data integrity (see the OpenZFS documentation on HW raid...with SW raid ontop it should be the same). Secondly adding more unneccessary storage/filesystem/virtualization layers adds more overhead. So a ZFS ontop of a mdadm would be slower than just a plain ZFS. So you should get a slower and less reliable storage.
Last edited:
cool, super easy!
it terms of performance is faster mdadm raid0 or zfs raid0 ?
I tried raid0 with mdadm and it was really fast
Its not all about performance. I for example would never use any kind of raid0 in a server except for maybe temporary data where loosing it isn't a problem. And as a enterprise product ZFS is designed for reliability, data integritiy and alot of great features like replication, snapshots, block level compression, deduplication and so on. This all adds alot of complexity and overhead so ZFS will never be as fast as a good HW raid or even a way more simple mdadm SW raid.
You can do thing as fast as possible risking failures or slowly and carefully double checking everything. ZFS is more the latter type.
Its the same like with overclocking. Overclocking will make your computer faster but you shouldn't overclock your server. A server should be as reliable as possible and not as fast as possible. Overclocking it for more performance would also increase the chance of hardware errors or even failures so no serious admin would do such a thing.
You can do thing as fast as possible risking failures or slowly and carefully double checking everything. ZFS is more the latter type.
Its the same like with overclocking. Overclocking will make your computer faster but you shouldn't overclock your server. A server should be as reliable as possible and not as fast as possible. Overclocking it for more performance would also increase the chance of hardware errors or even failures so no serious admin would do such a thing.
Last edited:
