Hi there,
When using ZFS, there is a recordsize. quts hero allows setting it up to 128kB. The recordsize is the maximum size one record can have. It is allowed to be smaller.
The main usage of it is to optimize “big file, random small accesses” and volumes for e.g. VMs. There is a performance benefit if the recordsize matches the commonly used access sizes. Read or write.
In a Z1 RAID of e.g. 4 disks, one record + its parity is split across the 4 disks. So each takes maximum 128/(4-1) kB units of data. This data can be sequential, but does not have to be.
The performs still ok for SSD storage, but HDD storages are struggling here.
In a scenario where you have files of multiple Megabytes, they are split in a lot of requests.
In addition, as 128kB/3 does not align with 4kB ashift of ZFS, there are a couple percent storage space lost. If the Z1 would contain 3 or 5 disks, this would not happen. Still, the loss scales down with higher record sizes as well.
It would be beneficial if the 128kB limit for user created ZFS shares would be increased. The ZFS used in 6.0 version already reports it can support 1MB recordsize. So why not leaving it up to the user to configure this?
Best regards
Florian