Testing read striping (Win 10)

[propergol]

Hello,

On my Windows 10 PC (insider build 10532 + DrivePool_2.2.0.636_x64_BETA) I did the following tests :

- 1 pool with 2 HDs, both HD are the same model : WD Green 4 Tb. Each drive tested outside the pool shows a read speed of  150Mb/s (empty disk)

- the pool has real time duplication, read sriping, bypass system filters set to ON and file protection set to file duplication.

- on this pool I have copied 2 movies (25Gb, 15Gb)

- on this pc one SSD capable of 470Mb/s write, and one HD 140Mb/s

 

Now...

- if I try to copy one movie from the pool to the SSD, DrivePool only read from 1 HD in the pool and so the copy is only 150Mb/s (while the SSD supports up to 470Mb/s)

- while the movie is being copied @ 150Mb/s to the SSD, I start to copy the second movie to the standalone HD : DrivePool shows that reads from the pool are only made on 1 HD and so copy to SSD drop down to 60Mb/s and copy to HD 60Mb/s  

 

 

 

 

I am very new to DrivePool, so maybe I miss something.

[propergol]

Thanks a LOT!

 

DrivePool 2.2.0.649 beta works and it is fast now 

I have no idea where the problem was but it is now fixed.

 

Now that read striping is fixed I would try to speed up the write thing with SSD plugin. I am waitting for some good deal (since I need 3) on fast 250Gb SSD.

 

Thanks again.

[Umfriend]

Drashna, thanks but I will wait for an official release before I update, I 'd rather not do beta's (although I would consider it for the String/Grouping function and FileSafe )

[Christopher (Drashna)]

Drashna. quick question:  During numerous rounds of observation of the read-stripping performance, I have noticed that it seems to only read from one disk, especially when the read-speed (which directly correlates to the write speed at the destination) is too slow to handle the additional bandwidth boost that reading from another drive simultaneously would provide. Basically, it seems to figure out that reading from more than one drive would not improve performance, as the outbound pipe appears to be saturated already, so why bother reading from more than one drive anyway. This is in the single large file copy scenario, not multiple smaller files to multiple destinations (e.g. streaming to multiple devices) scenario. This question was prompted by randomly checking the tooltip of the "Read Stripping" bar on the drive pool UI. In the large single file copy (~500 GiB file) scenario where I almost always only see reads from one drive in the X2 pool. The tooltip reports "65-70 per second Disk Holds" which makes me think that reading from another drive would not help anyway, so why bother. (It is a network file copy to a slow archive NAS that saturates at ~40 MB/s, slow parity protected array destination, unRAID box).  Is this assumption correct?  

 

I'll try to set up a RAM drive on the same box hosting the drive pool and set up some large file copies to the RAM drive, which should have *HUGE* write bandwidth and see if that makes a difference in what Read-Stripping does.

 

 

PS: StableBit should consider adding a RAM drive service to the suite of tools you already provide <nudge nudge> ;^)

 

 

Thanks a LOT!

 

DrivePool 2.2.0.649 beta works and it is fast now 

I have no idea where the problem was but it is now fixed.

 

Now that read striping is fixed I would try to speed up the write thing with SSD plugin. I am waitting for some good deal (since I need 3) on fast 250Gb SSD.

 

Thanks again.

 

 

The issue here was two fold. Both involved measuring the speed of Read I/O for the purpose of Read Striping. This didn't affect every system though.   

 

So it was getting "stuck" and would not read stripe properly, basically. 

 

 

@YahottaBKiddin:  You're definitely seeing the same issue then.  The linked version should fix your issue as well. If not, let us know.

 

As for a RAM Disk .... The biggest issue with RAM Disks is sudden power loss. You can lose a lot of data that way.

But I'll pass along the suggestion. 

 

Drashna, thanks but I will wait for an official release before I update, I 'd rather not do beta's (although I would consider it for the String/Grouping function and FileSafe )

 

Not a problem. We definitely understand that not everyone wants to use a beta build. 

Though, ... if you look at the change log, we do need to release a stable version soon, and I've been bugging Alex about that. It's been .... a year since a stable release for DrivePool (though, a lot of that has been that Alex has been putting a LOT of time into CloudDrive to get it where it is). 

 

And as soon as we have a grouping function in DrivePool (beta or otherwise), I'll post about it on the forums most likely (but you may notice a version increment (2.3).