Does Parity drive type influence overall performance

[seanant]

This may have come up before but want to ask now to see if there's any reason to install a high end parity drive or will a standard desktop drive work just as we'll?

[sureguy]

Most people use a desktop drives.  I'd make sure it's no slower than any of your data disks.

[garycase]

Basically the answer to your question is No ... but with a caveat.

 

A high-end parity drive won't make parity checks any faster, as they're limited by the slowest drive in play at any point in the check.    The only time a faster parity drive would help with this is if it was larger than all other drives ... in which case after the check had proceeded past the size of the largest data drive, the check would be faster, since all that would be required to complete it would be reading the rest of the parity drive.

 

But if you are using your server in a manner such that multiple writes are done at the same time, a higher speed parity drive will make those slightly faster ... not so much because of the higher spin rate, but because higher speed drives typically also have better seek times; so the average write speed will be improved a bit.

 

But in the vast majority of cases, it's best to simply use whatever drive you're using for your data disks.

 

[c3]

http://lime-technology.com/forum/index.php?topic=26140.0

 

Spoiler: The lower rpm drive yielded lower parity check time.

[garycase]

http://lime-technology.com/forum/index.php?topic=26140.0

 

Spoiler: The lower rpm drive yielded lower parity check time.

 

Not true.  Note the results:

Test Result #5 - Parity Check

    HH:MM:SS  MB/s  Parity  Disk1  Disk2

DX  07:59:00  139.2  44C    39C    39C

DM  08:40:00  128.2  34C    30C    31C

 

The lower rpm drive (DM) took 41 minutes longer -- NOT less.  BUT the faster drive ran significantly hotter (a normal result)

 

Note also that the entire test has virtually NO relationship to a typical UnRAID system, since all of the drives are the same.    Few folks are going to replace ALL of their drives when they upgrade their parity drive to 4TB 

 

The tester was also "surprised" that the slower rpm DM drives had faster data transfer speeds than the 7200rpm DX drives.  I wasn't surprised at all -- the DM's are 1TB/platter drives; the DX drives are 800GB/platter ... so the DM's have a 25% higher areal density, which easily compensates for the 18% slower spin rate.

 

As I noted before, there's NO significant advantage to using high rpm drives as parity.

 

[c3]

Oops, poor memory 

[seanant]

Very cool. Thanks.

[garycase]

You're welcome.    Note that not only is there no good reason to use high rpm drives for parity;  but that the lower rpm drives run cooler;  use less power;  and are likely (as a result of these factors) to last longer 

 

I'd use WD Reds for your parity drive if it's 2TB or 3TB.    They're not yet an option if you want a 4TB parity drive; so in those cases I'd use the 5900 rpm Seagates.    FYI WD is planning to release both 4TB and 5TB Reds later this year (I'd expect the 4TB ~ Oct and the 5TB in Dec/Jan).