Hi,

 

I think I've found a bug but I would like to verify it first.

 

Yesterday I started a parity check. unRAID calculated 1184 minutes and started the check. I didn't do anything with the machine and let it go.

 

10 minutes ago I had to read something from the array (no writes, definitely). When I looked at the main page the positions and estimated end time jumped significant from 584.xxx.xxx to 866.xxx.xxx (current position) and  321 to 39 (estimated finish).

 

Such a huge jump doesn't look good to me. Is it possible that Parity check takes the last read position (from my read - not the parity read) and go's on from that position?

 

Regards

Harald

 

Hi,

 

forget my previous post. Somewhere at this point my 750GB drives ended and the two 1TB drives (parity and one data) were left. Seems that this enhanced check performance significant.

 

Sorry.

 

Harald

 

Hi,

 

forget my previous post. Somewhere at this point my 750GB drives ended and the two 1TB drives (parity and one data) were left. Seems that this enhanced check performance significant.

 

Sorry.

 

Harald

 

That makes perfect sense.  Instead of having to read 12 (or more) disks in your array to calculate parity, the parity check process only has to read the one 1TB disk once it is past the capacity of the smaller 750GB drives.

 

Reading one disk is much faster than reading all of them, thus the huge increase in speed.

If your drives are connected to (through) the PCI bus, the speed differential is quite evident.  Reading from all the drives at once through the PCI bus bottleneck really slows things down!  As smaller drives on the PCI bus are completed, the PCI bus is not quite so crowded, and the speed is much faster.

 

Since I use the PCI bus for some of my drives, I notice this affect.  But I wonder if you had all your drives connected through high-speed buses, (jimwhite's name comes to mind) whether this slowdown effect still occurs (maybe to a lesser extent).

But I wonder if you had all your drives connected through high-speed buses, (jimwhite's name comes to mind) whether this slowdown effect still occurs (maybe to a lesser extent).

 

It definitely occurs with any set of mismatched drives, because drive access is so much slower toward the end of any drive, and parity check performance can not go faster than the slowest drive at any one point.  I have all SATA II, no PCI, and I see a large jump in performance when it finishes my 250GB's, then it tapers down again as it finishes the 320GB.  On finishing the 320GB, there's another huge jump, and then it slowly tapers down to the end of the 500GB's.

Thanks for all your posts.

 

As you could see in my "Need help" thread in the 4.2 board I'm mostly using 750GB Samsungs. In addition there's a 750GB WD and a 1TB WD. Parity is a 1TB Samsung. The Samsungs are much faster than the WDs (and cooler). The calculated finish time dropped by several hours once the check reached the 750GB barrier.

 

Sorry folks - I'm just a little bit nervous since my eSATA experience. I didn't expect to see so many red dots in the last several days ;-)

 

Regards

Harald

 

Harald,

 

Not to Hi-Jack this post, I was wondering are you seeing feedback from your Samsung drives and if so which model # driver are you using. I had purchased a Samsung in Lieu of a WD as the were quite a but cheaper on sale here a while back, A local company bought quite a few -- anyhow I could not get any feedback from the drive 735jb comes to mind as the model but not certain. Are they Quiet? The drives you have installed?

 

Thanks,

 

Dave

Harald,

 

Not to Hi-Jack this post, I was wondering are you seeing feedback from your Samsung drives and if so which model # driver are you using. I had purchased a Samsung in Lieu of a WD as the were quite a but cheaper on sale here a while back, A local company bought quite a few -- anyhow I could not get any feedback from the drive 735jb comes to mind as the model but not certain. Are they Quiet? The drives you have installed?

 

Thanks,

 

Dave

 

I love these Samsungs. They are quiet, stay cool, are fast and cheap. SMART was disabled for some but smartctl could enable it - no problem. After nearly 4 months with high read/write rates there's just one with 849 sector remapped. The other drives report 1 or 2 sectors - if any. None of the drives reports an error.

 

I started my array with Samsungs and swiitched to WD because of the low power consumtion. After one month or so I calculated the difference in price and power consumtion and now I'm back to the Samsungs ;-)

 

Regards

Harald

 

I love these Samsungs. They are quiet, stay cool, are fast and cheap. SMART was disabled for some but smartctl could enable it - no problem. After nearly 4 months with high read/write rates there's just one with 849 sector remapped. The other drives report 1 or 2 sectors - if any. None of the drives reports an error.

 

I started my array with Samsungs and swiitched to WD because of the low power consumtion. After one month or so I calculated the difference in price and power consumtion and now I'm back to the Samsungs ;-)

 

Regards

Harald

 

 

I bought 2 of them due to the price differential, got home and found an article showing them to use about 8 watts at idle and a bit more at load... compared to the WD that uses only 4.5watts at idle and about 8 watts at load. I'm thinking 3.5watts over 10 or more drives is a substantial saving. I agree though, the Samsung are just as quiet but run so much cooler than the WD drive I have.

 

I want to buy some more drives, and really want to use the WD for the energy savings, but the price is higher and the temps are a bit higher as well... decisions.

Sounds like there is some specsmanship going on here... if the drive consuming less power is hotter, then something is rotten in Denmark... 99% of the time, hotter means more power... there aren't any magical motors that are significantly higher in efficiency...

 

Jim

 

I've long suspected that Samsung may be 'cheating' a little, perhaps placing the temp sensor off in a cool corner of the drive case, while Seagate puts theirs in a hotspot, where it ought to be.  I can concur that my Samsung's are my coolest and fastest drives, probably quietest also.  But the average 10C difference seems a little too cool to be credible, hard to explain as you indicated.

 

It would be good to see some actual tests of the units, under comparable loads, checking various locations on the drive casings.

 

[Note:  But it should probably only be done with 'remote' temp sensing equipment, one of those hand-held aim-it-read-it units.  No expert here, but I don't believe you should be touching thermometers or other temp sensing units with metallic tips to a drive in operation!  Unless you really know what you are doing, the rule should probably be: no metal to metal contact, not even grounded, especially near the circuit board, not while the drive is running.]