I was reading a thread about using a raid controller to build a raid 0 parity drive to increase the parity drive speed.  Apparently it worked, using the Areca x8 controller and one other controller.  I have one of those Areca controllers, though I don't want to use it for this.  I do have a Promise Supertrak Ex8350 controller which I would be willing to use.  My idea is to take two drives on the RAID controller, raid 1, and make the resulting raid volume the parity drive for UnRaid.  While this shouldn't affect the speed one way or the other, it would make the system able to withstand the simultaneous loss of the data drive AND one of the parity drives.  I.e. it would be "raid 6 like".

 

I could then use the other 4 controller SATA ports as pass through.

 

Anyone have any thoughts on this idea?

 

jwc

This type of thought pattern has been mentioned multiple times.

The reality is, the parity drive is no more important then the data drives.

RAID1 on parity provides only a smaller amount of protection.

Having a warm spare available with the ability to rebuild a failed drive in a moments notice is a better alternative and does not require a RAID card.

 

However, using the Areca controller with the combined RAID0/RAID1 (SAFE) mode that I am using has it's benefits.

With a pair of sizable drives you can use RAID0 on the outter volume for parity and RAID1 on the inner volume for cache

 

This provides a slightly faster parity and a protected cache drive.

It's the cache drive that is unprotected in the current unRAID level.

 

I use 2 Seagate 1.5TB 7200 RPM 32MB Cache drives in RAID0/RAID1 safe mode.

This yeilds 2GB for Parity and 350MB for cache.

On my cache drive I also have some hidden directories that I use for /home mounts via NFS from other systems.

 

This provides an always online nfs area and spinning parity for any writes to the most commonly used drives in my array.

 

I've tested hardware raid with Areca (both the X8 and X1 cards) and a 3WARE/AMCC PCI-X 4 port SATA card.

The 3WARE does not do the raided volumes. Each drive belongs to a full array.

The Areca lets you carve up a set of drives into various RAID levels.

It also caches some data so it provides an improvement in write speed, A UPS should be mandatory if you are using the cache.

 

I have not read that the Promise Supertrack is supported in a raid configuration.

It was your post about the raid 0 parity that started me thinking about this. 

 

I don't understand your comment that "the parity drive is no more important then the data drives".  I don't really understand the behind the scenes stuff but the way I understand it is that the parity for any given drive is stored two places, on the specific data drive and the parity drive, IOW parity info for Data Drive A is store only on Data Drive A and the parity drive.  If that is the case then if the parity drive is a RAID 1 array then you could lose the data drive and one of the two parity drives and still recover, whereas in the straight UnRaid, if you lose those two drives (Data Drive A and Parity) you have lost the contents of Data Drive A.

 

Thus my comment about "RAID 6 like".  In raid 6, if you lose any two drives you can still recover.

 

 

Extra protection of parity is always just that an extra level of protection.

Chances are the parity drive is a recent drive vs the data drives that may have been in service before unRAID.

 

In any case, if a drive fails, the next required action is getting a replacement drive to fill in the gap.

Would you rather have 2 parity drives spinning while awaiting the delivery of a drive to replace the one that has failed?

 

If I have a 15 drive array with RAID 1 on parity and a drive fails. then all 17 drives must remain spinning if writes occur to the missing drive. Simply mounting or un mounting the drive requires a write to the missing drive.

 

I guess it's a gamble, the chances of loosing a parity drive (Which is probably more recent) vs the time it takes to get a replacement drive installed and working.

 

To me, Having a replacement readily available to fill in the gap is more important then duplicating the parity drive.

 

If the parity drive fails. You replace it. if a data drive fails you replace it.

 

Consider this, if you have a double drive failure, what are the chances it is one of the more recent parity drives vs one of the older data drives.

 

If I had a spindle of data that had critical information, then putting that on a raid1 array would buy me extra piece of mind.  This is what I do with my cache drive. RAID1 and hidden directories of my most used files in home directories.  These directories are also rsync'ed onto one of the array drives on frequent basis.

OK, I understand your argument.

 

I went out and read the whole parity thing and obviously my understanding of how this stuff was "stored" was just way off base.

 

jwc