Ah yes, I see.  It seems like having multiple arrays would be more useful than having an extra parity drive... 

For a system with 20+ drives in a single unRAID array it seems very risky to have single drive redundancy. Many posts talked about how to go about solving this (mirroring parity drives i.e.) - the most important thing is not to forget what the needed feature is and that is especially for arrays with more than 10 HDDs to have multiple drive redundancy. Isn´t this the whole idea of unRAID to be able to build efficient arrays with affordable equipment? The MTBF of 2 drives with more than 20 drives is quite low!

 

 

For a system with 20+ drives in a single unRAID array it seems very risky to have single drive redundancy. Many posts talked about how to go about solving this (mirroring parity drives i.e.) - the most important thing is not to forget what the needed feature is and that is especially for arrays with more than 10 HDDs to have multiple drive redundancy. Isn´t this the whole idea of unRAID to be able to build efficient arrays with affordable equipment? The MTBF of 2 drives with more than 20 drives is quite low!

 

 

 

This is balanced by :

 

- bunches of your array drives being spun down for a good percentage of the time and not in use 24/7 *potentially* extending the MTBF

 

- unraid only being vulnerable to losing data on failed drives (beyond 1 failure) and not losing data on the whole array in that event ala striped raid implementations.

 

This lessens your exposure dramatically and, in my eyes, makes it less essential to have multiple parity than it is under a striped system.

 

It would still be a nice feature but I'm not sure how pressing it actually is.

I guess I just don't see how you gain that much additional protection by having a second parity drive installed in a single array.  It really only protects you against a failure of the parity drive while attempting to replace a single failed data drive.  Now if it's a different type of parity calculation that allows you to recover from two simultaneous data drive failures then that is a benefit.  However it was previously mentioned in this thread that that second type of parity calculation would come with a performance penalty.  Splitting a large array in a single server into two smaller arrays in the same server then that might be useful as well.

Not sure this is clear in this thread, but the idea of "dual parity" is not to split the array into 2 arrays in the same box.  The term "dual parity" is not even technically correct, just a convenient way to refer to the idea of using 2 drives for data redundnacy.  One would be parity.  The other would use a complex mathematical technique to provide the ability to recover a second failed disks.

 

Refer to this thread.

 

You might also want to use Google to read about RAID6, which uses a similar mathematical technique.