plugout.net

Hello lovely devs,
I have a great idea! Basically, I was looking at a way to use all of the space from all drives of dissimilar size and OpenZFS seems to be the only open sourcy way to mix and match sizes. Perhaps it's worth a look? I'm mainly thinking of the Storage Pool idea, it's redundant but uses all of the space of random-sized drives. For instance, I have 2 2TB drives and 2 1TB drives in a RAID5. I can use 2.7TB right now, but with storage pools I could use around 3.5TB etc.

http://zfsonlinux.org/
https://github.com/zfsonlinux/zfs/

"Native ZFS for Linux!

ZFS is an advanced file system and volume manager which was originally developed for Solaris and is now maintained by the Illumos community.

ZFS on Linux, which is also known as ZoL, is currently feature complete. It includes fully functional and stable SPA, DMU, ZVOL, and ZPL layers."
-------------------------
With the newer 4.3.3 kernel, it may be possible to use BRTFS instead and that has a better license...

---Added BRTFS to the previous message---

Hi andjohn,

I am sorry but I don't see benefits to use BRTFS with lacie.

Yes, this fs provides better safety controls (snapshots and checksum...) but stays slower I/O access than an ext fs. So as LACIE NAS don't have the best performances about it, use it on data partition does not seem to be a good idea.

For the other partitions, it doesn't worth to use it as we can easily restore the system partitions

Yes, I agree about the other partitions, but I'm mainly looking at replicating the feature of the Lacie stock firmware where you can mix and match hard drive sizes like a Drobo and still use all of the space. That's what both of these styles can do.

:shocked :shocked :shocked
On a data disk, unused space are the orphan system partitions (<=> sdx1,2,3,5,6 and 7) so 2,7GB (4*512MB + 8 + 768MB) <=> a video (High resolution)

so for example with a 1TB your challenge is to not to waste 0.27% of disk size (0.13% on a 2TB; 0.08% on a 3TB) :dry

For an average price of 50€ /TB, 0.27% <=> you spent 0.13€ for this unused space.

I see what you're saying, perhaps I wasn't more clear. I have 2x 2TB drives and 2x 1TB drives that I have in a RAID5. Currently one of my 2TB is my primary drive, but I've had to make my array 4 times the lowest size, or around 1TB per disk for a total near 4TB.

If I was able to use the entire disks, it'd be more like 2x 1TB + 2x a bit less then 2TB or around 6TB for the array. Also, it would be easier to expand if I bought a 3TB for instance to replace one of the 1TB drive. That's mainly my aim, easy expandability and using all of the disk space for parity and storage.

Ok I understand better what you mean.

andjohn wrote:I see what you're saying, perhaps I wasn't more clear. I have 2x 2TB drives and 2x 1TB drives that I have in a RAID5. Currently one of my 2TB is my primary drive, but I've had to make my array 4 times the lowest size, or around 1TB per disk for a total near 4TB.

:disapprove Never you can have 4TB raid size with 4 disks of 1TB with a raid5.
the raid5 with 4 x 1TB devices is = (4 - 1)*1TB => 3TB (at each writing, a member is used by redundancy).

What you may do is:
to build a raid0 with the 2 1TB disks to get a 2TB raid0 and include it in a raid 5 with the 2 others disks
Then raid5 size will be : (3-1)*2TB => 4 TB

That's true, I wasn't including parity in any of my calculations. I still think BRTFS would be advantageous.

brtfs is supported by the kernel (3.1.4.2 and the experimental 4.3.3) you only need to load the kernel module
Use at own risk, I read on the Internet some issue with this file system

Thanks guys, i could be a bit early for BRTFS. Perhaps I'll wait until its a default in a distro...