I am a little confused by your expectations; ZFS is a filesystem indeed, but BeyondRAID is a RAID implementation, not a filesystem, as far as I can tell (the Drobo probably has its own filesystem implementation, but I do think when you speak of BeyondRAID you are comparing it to RAID-Z, not ZFS). Perhaps what you mean to say is “RAID-Z is the only better RAID implementation than BeyondRAID”?
Here I would like to point out that Linux has its own RAID implementation in the form of mdadm, and it is already at a rather mature stage. When used with LVM2, the logical volume management module, it easily matches ZFS/RAID-Z feature-wise, although it is still susceptible to the
RAID-5 write hole.
One critical thing to note about ZFS, as of the time of writing, is that it does not support RAID expansion. What this means is that if you already have a 4-disk RAID-Z/RAID-Z2 array, you
cannot expand it at a later date by adding one more disk to the array. This was one of the critical flaws of ZFS (for a home user; let us not speak of enterprise usage here) that led me to decide on LVM2 instead (yes, mdadm/lvm2 on linux allows you to do RAID expansion).
I had been keeping some tabs on development in this area; last I read is that
Adam managed to come up with a sketch for implementing this. But now that he and many other ex-Sun engineers
have left Oracle for other projects, I find it highly unlikely that Oracle will ever add this feature to ZFS, considering the lack of enterprise interest in this area (enterprises often expand their storage pools by adding
an entire array, rather than expanding existing ones. Such a feature is low on their priority list).
If you ever decide to expand your RAID-Z array, you are going to have to back up all data on it first, and rebuild the array from scratch. Not a minor undertaking for a non-enterprise user by any means if you have data running in the terabytes, and no spare disks.
It just occurred to me that there is no cheap motherboard (~$100) in existence (that I know of) that has at least nine SATA ports. Higher-end motherboards have up to ten (with a secondary on-board controller), but those are often meant for higher-power systems (X58 chipsets and the like), which runs against your idea of a low-power file server.
Even with such a ten-port SATA system, there is something to be said for not putting all your disks on the southbridge; apart from the AMD 890GX and Nvidia nForce series, neither of which are particularly low-power, pretty much all southbridge interconnects have a maximum bi-directional bandwidth of 1GB/s, and I have not heard of any consumer-level southbridge managing over 700MB/s. If you are going to try to implement FireWire and/or gigabit LAN connectivity, that southbridge interconnect is going to get very crowded. Even getting a cheap four-port SATA controller on the northbridge PCIe slots would alleviate that load quite a bit.
A cheap motherboard with a decent southbridge (ICH10 or maybe SB750+), plus an additional disk controller (RAID or otherwise), is a cheaper and more power-efficient solution. Here I readily assume you will be running the file server with onboard graphics.
