Can anyone explain the 'disparity' between the read performance of an unRAID array with and without a parity drive, in this SmallNetBuilder review of unRAID?  See the first graph (Figure 10), and the comment above it.

I'd guess they installed the parity drive, and then did their tests before parity calc was finished. 

For their sakes, I really hope that isn't true, rather embarrassing for the reviewer/tester.

 

Since most of us have no plans to drop our parity drives to test this, has anyone done comparable testing?  Or is anyone in a position to try testing this?  As far as I know, a parity drive should not have any impact at all on read performance.  Unless of course, the parity drive is in use, for concurrent needs!

I run my 2-drive test system currently with just 1 data an a cache disk.  I can make the cache disk parity, and do some testing.