Quote:Why pay $300 compared to this motherboard. The motherboard is unavailable from newegg but is available other places for $70. It has 6 SATA slots available and this guy uses it in all his setups.
This motherboard has 6 onboard SATA plus the onboard LSI controller adds 8 more ports on a PCIe 8x controller. As well by using a HP expander card thoes 8 ports expand into 48. This cannot be done with a standard onboard controller. The HP expander card costs ~150$.
Thats very cheap/port when needing many hard drives
ZFS vs other filesystems is a big topic. Major points are
-ZFS is a copy on write filesystem and every block is checksumed. That means if a file is corrupted when you read it back the filesystem knows this and recovers automatically. This can detect phantom writes, misdirected reads and writes, DMA parity errors, driver bugs and accidental overwrites as well as traditional "bit rot."
-Free space is pooled across all filesystems. For example you could have 10tb of free space, and allocate 5x 10tb filesystems on this. each 10tb filesystem only uses the space that it requires despite allocating 10tb
-Built in compression (and its fast!)
-ZFS is a filesystem, raid system, and disk manager. Everything is encompassed by the ZFS tools. No LVM, no Raid layer
ZFS also does not have the write hole issue so expensive battery backup raid controllers are not needed! If you loose power when running raid you can have parity/data mismatch issues and recover corrupt data. ZFS bypasses this. ZFS loves cheap disks and controllers!
-de-duplication. If two pieces of data are identical on the stoarge pool they only take the space required for one. This can be done at the file or block level. (needs a alot of ram though, this feature is also bleeding edge. Enabled only in the very latest solaris and freebsd kernel)
-Built in NFSv4 ACL, iSCSI* and samba* sharing (* = solaris kernel only. freebsd or linux can use 3rd party tools)
The major downsides to ZFS are;
- Its maintained by oracle now. (Booooo Oracle!)
- You cannot grow Raid-Z1 or Raid-Z2 arrays. RaidZ is like raid5/6 but without the write hole issues. You can add another whole array to your pool to grow it, but you cannot add more disks to an already created array
- No linux kernel driver (yet). There is a userspace FUSE driver, but its slower. Some guys from India are working on it though. I run mine on freebsd, which is awesome and my favorite server OS by far.
- Some performance hits but still very fast. my 7 disk 1tb array in raid-z1 reads ~330mb/sec writes ~150mb/sec.
