I just did some q&d testing comparing modulo with division and bitwise AND, and the differences were not worth quibbling over.

 

Bitwise ops were faster, but only a factor of 4 in the worst case.  (however I don't have  profiler, and was using clock() which is not really high enough resolution.

 

Interestingly, the modulo function appears to be highly optimized, as x%29 took two to three times longer than x%32.

 

But all were orders of magnitude faster than XORing two 512byte sectors.... so it doesn't seem to be a critical issue.

It makes sense that division by 32 (and the remainder from dividing by 32) would be fast. It's just a shift operation, right?  I would think modern compilers would catch that.  A shift on a desktop-class CPU is a 1-cycle operation, isn't?

Any new news on dual parity?

Funny, it just happened to me. Well I replaced one disk, and while replacing my parity drive with a larger disk, I noticed one of the 750GB small disks was starting to fail. Yikes! You know, you could do a hardware raid 1 on the parity drive. I think most motherboards today offer that option. I am not sure if it will be seen by unraid correctly. I know I had a raid 0 that worked fine (just testing). Never thought of trying a raid 1 for parity. Of course the problem I had with the raid 0 is that it never powered off. This was over 1.5 years ago though, on older code. Who know today what the different would be.

Funny, it just happened to me. Well I replaced one disk, and while replacing my parity drive with a larger disk, I noticed one of the 750GB small disks was starting to fail. Yikes! You know, you could do a hardware raid 1 on the parity drive. I think most motherboards today offer that option. I am not sure if it will be seen by unraid correctly. I know I had a raid 0 that worked fine (just testing). Never thought of trying a raid 1 for parity. Of course the problem I had with the raid 0 is that it never powered off. This was over 1.5 years ago though, on older code. Who know today what the different would be.

 

Trying to spin down the RAID0 set would still fail if I remember correctly.

 

The RAID1 might spin down.  If unRAID thinks there is only one drive and a spindown command is sent to it the other drive might spin down, though I think you are going to have the set the spindown timer on the drive itself before it will work the way you would like it to.

Funny, it just happened to me. Well I replaced one disk, and while replacing my parity drive with a larger disk, I noticed one of the 750GB small disks was starting to fail. Yikes! You know, you could do a hardware raid 1 on the parity drive. I think most motherboards today offer that option. I am not sure if it will be seen by unraid correctly. I know I had a raid 0 that worked fine (just testing). Never thought of trying a raid 1 for parity. Of course the problem I had with the raid 0 is that it never powered off. This was over 1.5 years ago though, on older code. Who know today what the different would be.

A RAID1 parity would not have helped you when upgrading the parity drive.  You would have had to replace 2 parity drives with larger ones AND the smaller 750GB drive would still have been failing.

 

Any hardware RAID of the PARITY drive is useless in my opinion.  The parity drive is absolutely no more, or less important, than every other disk in your server.  The only benefit would be to "possibly" increase write speeds if multiple data disks were being simultaneously written to.  ("possibly" because the writes on different data disks might still end up with stripes on the same parity disk.)

 

A true diagonal parity or reed-soloman drive would permit you to recover from a pair of failed drives.  It would greatly complicate the upgrade/expansion of an array.

Funny, it just happened to me. Well I replaced one disk, and while replacing my parity drive with a larger disk, I noticed one of the 750GB small disks was starting to fail. Yikes! You know, you could do a hardware raid 1 on the parity drive. I think most motherboards today offer that option. I am not sure if it will be seen by unraid correctly. I know I had a raid 0 that worked fine (just testing). Never thought of trying a raid 1 for parity. Of course the problem I had with the raid 0 is that it never powered off. This was over 1.5 years ago though, on older code. Who know today what the different would be.

A RAID1 parity would not have helped you when upgrading the parity drive.  You would have had to replace 2 parity drives with larger ones AND the smaller 750GB drive would still have been failing.

 

Any hardware RAID of the PARITY drive is useless in my opinion.  The parity drive is absolutely no more, or less important, than every other disk in your server.  

 

Let me clarify that I agree that any hardware RAID1 of the PARITY drive is almost useless.

As Joe mentions, The parity drive is no more important then any other drive in your system.

I won't say less, because it is slightly less important then any data drive.

If you loose the parity drive, your data is still 100% intact (just unprotected).

 

Are there any unRAID statistics or claims where a double drive failure occurred where the second failure was the parity drive?

This is the only case that RAID1 on parity will protect you.

 

The only benefit would be to "possibly" increase write speeds if multiple data disks were being simultaneously written to.  ("possibly" because the writes on different data disks might still end up with stripes on the same parity disk.)

 

RAID0 on PARITY Does increase write speeds, but not significantly enough that people will run out and buy RAID controllers.

I'll add that RAID0 PARITY/RAID1 CACHE is a high benefit factor to using the Areca hardware raid controller and fills a gap where no protection currently exists. I.E. The cache drive.

 

Adding to some of the comments below.

When using the Areca RAID controller, in RAID1, RAID0 or Combi RAID0/RAID1 Safe mode. The drives spin down.

There is a timer in the bios that lets the drive spin down when there is no activity for 60 minutes.

 

Because my "SAFE" Arrangement has 2 virtual drives tied to 2 phsyical drives. I wrote a cron job to keep the drives active between the hours of 6:00 and 23:00 so I do not have to wait for them to spin up. Outside of those hours the RAID drives do spin down.

 

 

A true diagonal parity or reed-soloman drive would permit you to recover from a pair of failed drives.  It would greatly complicate the upgrade/expansion of an array.

 

This is what unRAID needs to allow safeguarding a larger set of drives (or having multiple arrays).