multiNAS (2 computer) concatenation? (ie 2 box, 4TB each, = JBOD 8TB?)

[mediahound]

My big posts got too big so i'm changing to single small questions now since they are critical questions changing my strategy of what to build/how to communicate what I want.

Is there any way to do the equivalent of BIG/SPAN/JBOD type drive concatenation, but over a network, two smaller NAS boxes presented as one larger drive to samba/NFS/etc?

If I had 2 physically separate motherboards and NAS boxes, lets say they have 4tb each (each with dual 2tb drives), and they are on the same network... is there any way I could make them act as a single 8TB drive?

For instance, could a single Zpool somehow access a 2nd NAS box to include it's drives? I'm aware performance would be an issue at 1gig ethernet speeds but i'm just curious whats possible since when 10gig ethernet becomes supported and less expensive it may be doable.

Another possibility would there be some kind of software that would let me mount multiple NAS boxes to treat them as a single concatenated drives from within freenas or linux?

 

[Frangen]

My big posts got too big so i'm changing to single small questions now since they are critical questions changing my strategy of what to build/how to communicate what I want.

Is there any way to do the equivalent of BIG/SPAN/JBOD type drive concatenation, but over a network, two smaller NAS boxes presented as one larger drive to samba/NFS/etc?

If I had 2 physically separate motherboards and NAS boxes, lets say they have 4tb each (each with dual 2tb drives), and they are on the same network... is there any way I could make them act as a single 8TB drive?

For instance, could a single Zpool somehow access a 2nd NAS box to include it's drives? I'm aware performance would be an issue at 1gig ethernet speeds but i'm just curious whats possible since when 10gig ethernet becomes supported and less expensive it may be doable.

Another possibility would there be some kind of software that would let me mount multiple NAS boxes to treat them as a single concatenated drives from within freenas or linux?

I think you can use iSCSI to publish the free space on the VDev from one of the nas boxes. Then import this as a disk or two disks (depending on your raid setup) to your main nas server.
This way you can stripe it with an existing VDev or create a new VDev (raid) from scratch.

P.S
I am new to freenas, someone else could probably explain it better.

 

[Caesar]

I'm not sure why you would want to do this but I think when you use iSCSI you have the same access to the iSCSI drive as an internal drive so maybe you can iSCSI from both servers and then let your OS JBOD the drives? I am new too so I am not sure if this is possible or not.

 

[Frangen]

Well if you have same problem as me...

I do not have any more room for hard drives on my freenas box. But i would like to keep adding more space to my current zpool.
This is how you could do it:

My freenas is setup with 6 drives in a 3x2 raidz cofiguration. (This would be similar to raid 50)
But now all my drives are full, and i need to add more storage space. This can be solved by installing freenas on a second machine, insert 3 drives and publish each drive as a single iscsi device.
I can then import these to my "main" freenas box as they where locally mounted drives, and add them to my zpool.
Keep in mind that raidz is only as fast as the slowest disk, so this type of configuration could have very slow write speeds.

Configuring iscsi is a bit tricky the first time., i have only done this on my esxi, never on freenas.

http://www.youtube.com/watch?v=ANgFxyfyY_Q

 

[Frangen]

If you whant to use jbod, just publish disks as iscsi , and stripe them together with the drives on your main freenas.

 

[ProtoSD]

It sounds like you're looking for something like this:

Tahoe-LAFS

I thought there was a feature request for this awhile back, but it may have been something similar.

 

[mediahound]

My real reason for "wanting to do this" was receiving a negative reaction to my serious questions about needing to build a petabyte array. >_> But also needing to clearly experiment on a much smaller system before I try something too grandiose betting the farm on it.

One reason is wondering whether it might work around a RAM bottleneck - even if you can physically mount the drives it sounds like ZFS wants about 1gig ram per TB of drive space, whether if there's a way to have "32gigs ram, 32terabytes hard drive" per unit, yet present that as a single 64tb (or more in the future) drive to outside access, yet each segment of ZFS just manages it's own drives somehow if that makes sense... though when i think about it mounting remote drives as local you'd still need the ram to map out their information so it's probably only useful where running out of drives was the problem. So it might have to be something separate from or working over the top of ZFS... maybe ZFS needs a way to distribute drive requests intelligently or just sending the same request for data to both servers with each half knowing it's only responsible for dealing with half the data? :P

Tahoe LAFS looks interesting... can that be used with ZFS?