Page 1 of 1

Speed & split images for recovery

Posted: Tue Aug 20, 2013 5:10 am
by GeraldEurope
Hi,

I'm a new user who licensed yesterday and I am very very pleased with this software. It is currently helping me recover all data from a RAID5 from a Sharkoon 5-Bay RAID box where one HDD died and another one is OK but is for some reason not detected ok by the box. I found the parameters and built the virtual RAID5 and all is working fine.

Now I have 2 questions:
While all works fine and speed is not really bad I wonder if there are some ways to improve it further? It's recovering at a rate of around 5-10 MB/s from 4 pretty new 3,5" HDDs (connected via SATA on mainboard directly) and writing to another one, also rather new 3,5" HDD (I even tried restoring to a Crucial M4 SSD and wasn't able to get over 10 MB/s). One HDD (died) is missing from the original 5-HDD RAID5, so R-Studio also has to use the PD to restore the data. Is that the reason why it's "only" at 5-10 MB/s?
Another reason I can think of is this: the RAID5 parameters I found have a block size of only 512 bytes, so maybe this is causing a lot of "very tiny" read and write operations, which is causing the lower speed.

Please don't misunderstand this as criticism that it is "slow", I guess this is already pretty fast for a data recovery tool, I just wonder if it could be faster or some optimizations either in my hardware setup for restoring, or in the R-Studio code (better caching for these 512 byte read/write operations?) are possible?

What I am also trying now, and which is my second question:
I am trying to create images of the HDDs and then restore from those images (maybe this will be faster? but in any case this would give me more time for restoring as I want the original HDDs go back to the RAID box as soon as possible). Now I almost have enough space for all 4 HDD images, but for one I probably won't find the space, at least not without buying a new HDD... For the last one it would be possible however to split the image (basically the image file will be around 1.1 TB and 600 GB could go to one HDD and 500 GB to the other one).
Now I wonder is it possible to load the image file in R-Studio and do a restore from the image file if it is split across 2 HDDs (ie the first part with 600 GB is stored on temporary backup HDD A and the other 500 GB part on temporary backup HDD B).

Thank you and thanks for this great software!

Re: Speed & split images for recovery

Posted: Tue Aug 20, 2013 2:13 pm
by Alt
GeraldEurope wrote: Another reason I can think of is this: the RAID5 parameters I found have a block size of only 512 bytes, so maybe this is causing a lot of "very tiny" read and write operations, which is causing the lower speed.
I think this is the most probable cause, but R-Studio doesn't read disks, Windows does. Cashing won't do anything, as R-Studio reads disks directly, bypassing disk caches.
GeraldEurope wrote: What I am also trying now, and which is my second question:
I am trying to create images of the HDDs and then restore from those images (maybe this will be faster? but in any case this would give me more time for restoring as I want the original HDDs go back to the RAID box as soon as possible). Now I almost have enough space for all 4 HDD images, but for one I probably won't find the space, at least not without buying a new HDD... For the last one it would be possible however to split the image (basically the image file will be around 1.1 TB and 600 GB could go to one HDD and 500 GB to the other one).
Now I wonder is it possible to load the image file in R-Studio and do a restore from the image file if it is split across 2 HDDs (ie the first part with 600 GB is stored on temporary backup HDD A and the other 500 GB part on temporary backup HDD B).
R-Studio can read images split into several files but only when all the image files are on the same folder, not to say about disks. Maybe, if that is absolutely necessary, it will be possible to combine the image parts into a compound volume set and include that volume into a virtual RAID. But I never used this scheme and I'm not sure it'll work, especially for compressed images.