dual parity and dual cache......why?

[French-Guy]

What minimum CPU would be necessary to run a dual parity disk in good conditions?

 

[Frank1940]

If you are still running that sempron, the parity check time will  approximately double.  (I had a dual parity setup on my test bed server -- running a sempron 140 back when dual parity was first introduced.) If my memory is not playing tricks with me, I seem to recall that Intel chips preformed a bit better than AMD.  I also seem to recall that a CPU had to have around 2000 passmarks to eliminate the time differential between dual and single parity.  Be a bit careful as the passmarks available for each core is a factor as I believe that the parity operation is single threaded.  

[French-Guy]

I didn't purchase that FX-8320e finally.............was too busy at work, and couldn't go to my Microcenter on time to get it for $40 !

Anyway, I have a AMD X2 250 with 4GB memory doing nothing, so i will do some tests.  It won't be a huge improvement of course, but still better than my current Sempron

Will see how the parity check improves on my current server (just 1 parity drive, but it took a looooooong time with my Sempron !!!)

And I'll keep an eye on some good deals for Christmas 

[JorgeB]

56 minutes ago, French-Guy said:

What minimum CPU would be necessary to run a dual parity disk in good conditions?

 

Any Sandybridge or newer Celeron/Pentium is enough for a small/medium array.

[French-Guy]

Was just looking at Microcenter for their weekly ads, and saw this:

AMD FX 8320e + ASUS M5A78L-M Plus/USB3 AM3+   =   $50 after rebates !!!

[dchamb]

1 hour ago, Frank1940 said:

 

No.  Parity 1 is a very simple binary operation.  Parity 2 is a complex matrix operation and is very CPU intensive UNLESS you have a newer CPU which has an new operator which reduces the number of CPU cycles to do the calculations.

 

 

You can ANY two drives fail and not lose data.  That includes both parity and data drives.  So you can have two data drives fail, a data and parity drive, or a two parity drives fail without data loss.  But if a third drive fails with any of these combinations, you will be losing data! 

Thanks Frank1490,

 

Is there a document somewhere that describes how the second parity algorithm works?

 

My unRAID server uses an Intel i5-3570. I'm pretty sure I'm covered.

 

Dale

[Frank1940]

8 minutes ago, dchamb said:

 

Is there a document somewhere that describes how the second parity algorithm works?

 

Start here:

      https://forums.lime-technology.com/topic/49598-unraid-server-release-620-rc4-available/#comment-487759

 

 

and google     Q parity calculation     and the papers that you find are not for the faint of heart.  The theory behind the whole thing involves a lot of matrix algebra.  (I hold a BSEE degree from many, many years ago and I when I was looking at the mathematical basis for this approach I soon realize how little I remember from my college years!) 

[bonienl]

If you are in mathematics, see this paper.

 

p+q, raid6.pdf

[pwm]

The math for second/third/fourth/... parity is normally only covered in quite high-end math-specific courses.

 

Anyway - if you think it was slow to compute the parity for the first drive, then it was because of the capability of the machine to read/write all the data.  Even slow processors can compute XOR at a number of GB/second so it's the disk transfer rates that decides the time.

When adding additional parity disks, then the CPU time for parity computation starts to affect the total parity build time.

 

The algorithms to compute parity is similar to what is used in normal software RAID on Linux or what is used by Snapraid etc. And it's the same algorithms as computed by dedicated RAID controller cards.

[French-Guy]

That's funny........Did you know that Galois was a French mathematician, born in 1811???

 

[pwm]

3 minutes ago, French-Guy said:

That's funny........Did you know that Galois was a French mathematician, born in 1811???

 

Yes, but then quite few people learns enough math to reach into 19th and 20th century math.

 

There are a few exceptions where some reasonably simple concepts was discovered rather late.