XigmaNAS

Hello. I am looking for the experts, as I am not having easy time getting an answer from someone with experience.

I have currently 8x 2TB drive on a LSI/Dell perc 6i
I built a hardware Raid6 as I was sold on adding capacity to array after creation, no problem!

Well added drives are indeed added to array. Adding 2tb drives x2 took about 96 hours.

Problem is you cannot increase the size of the volume, but only add another logical. So I have large volume "D" but now small volume "E" which is the extra, added capacity.

Thank you for reading this so far!

ZFS Z3 supposedly allows you to use drives of mismatched sizes, and also add capacity at any time.

1. Had anyone used the hodgepodge ZFS array and what is your feeling on it? I am not looking for speed, but data preservation, so Z3 is needed.
It would be nice to add some 1tb, then if price drops, a 2tb and so on.

2. Adding capacity on the go, is it much faster than windows resize times in zfs? adding 4tb capacity took 96 hours.

3. Is it wise to keep extra motherboard (same bios, same revision) in case of failure, or is it TRULY safe the zfs ability to
add all drives to new MB, load the array file (sorry I not know the name) and all is ok?

Many thanks for your very patient replies.

raidz2 is the equivalent of raid6

here is a model of the expected failure rate of drives in those configurations
http://www.servethehome.com/raid-calcul ... tdl-model/

yes you can use mismatched sizes, but the smallest drive determines the size of all the drives.
so 6x1tb + 1x120gb = 7x 120gb for the array
but what you can do is later replace the 120gb drive with a 1tb drive and the array becomes a 7x1tb array.

what you can do however with zfs is add another vdev to your pool.
in that situation raid z3 would be a bit painfull.

what you could do for example is create vdev mirrors for 2 drive each and keep adding those mirrors to your pool.
in that scenario you would lose about 50% of your storage capacity
and it would only take about as much time as you would need to write the commands for this.
you could do the same with raidz3 but you need at least 4 drives to create a raidz3 vdev.
so to add one raidz3 drive you would need to add at least 4 drives each time.

more about zfs
http://forums.freenas.org/index.php?thr ... oobs.7775/

as a side node. raid storage expansion is very dangerous for your data. if the machine would have died during the expansion, you would have lose all your data.

3) normally you can just plug the drive into another machine and import the pool.
however if your perc 6i does not pass the drives natively to the os and you have to create a jbod/single disk raid then you will have a problem if the controller dies.

also if data preservation is important you will need ecc memory
http://forums.freenas.org/index.php?thr ... zfs.15449/

My concern is with many vdev, any transfer from once volume to another is incredibly slow.

On dual volume perc6i, transfer from one volume to the other is very slow, now add in the truecrypt it is perhaps 15-17mb / sec transfer slow transfer of many small files.
When reading or copy over network (not on same controller) it is fast fast.

3) Pass natively?

what do you mean by transfert from one volume?
you cant transfer files betweens vdevs that are in the same pool, because a pool is just one filesystem.

using encryption in nas4free means encrypting the block device, for that you want a cpu with aes-ni support and you will hardly feel the impact.

3) meaning it mode
http://forums.freenas.org/index.php?thr ... owd.11901/

b0ssman wrote:what do you mean by transfert from one volume?

Ah ok sorry for unclear english...

In the current raid array, because I added drives it has two volumes (d,E)

Both are encoded with truecrypt, each has separate key. So any transfer from D-->E requires TC to decode, re-encode.
Also data is being moved about on the same raid card. So it's very very slow as data must move drives-->card-->memory-->cpu (decode, rencode)-->memory-->card-->drives

I hope you understand?

I am hoping with ZFS it will be faster as parity method is different than dual stripe raid 6