Re: [Bug-ddrescue] Bug-ddrescue Digest, Vol 102, Issue 4

[Scott Dwyer]

On 7/7/2014 12:24 PM, William wrote:
> It's painfully slow. I already tried that way. It is impractical to 
> use "--skip-size=0" in damaged hard drives.
Correct. The only way to get the most good data first is to do multiple 
passes with skipping. There is no way to go fast in one pass without 
missing lots of good data. Increasing the skip size can actually very 
much improve getting the most good data first. The benefit of the 
reverse pass in this case is that it will get the good data that was 
skipped while going forward (because it comes from the other direction 
it will pick up on the tail end of the skipped areas). Simply running 
the second pass in the same direction will be much less likely to pick 
up all of the possible good data.

I will try to give an example of why the reverse pass is helpful, 
especially with a good skip size. Below is a (very) simple pattern of 
reads, each number represents a drive head. Head 2 is the bad one. The 
"S" below represents skipped data. So in the first pass forward we end 
up skipping part way into head 3, which is good readable data. If the 
second pass is also forward, we can still miss some of that same data 
from head 3 and have to wait for the slow no-skip pass to get it. But if 
the second pass is in reverse, we get the data from head 3 and skip into 
head 1, and we already got the data from head 1 on the first pass. The 
only downside is that the reverse reads are slower. But that is a small 
price to pay for getting the best data first. The trick is to find the 
most effective skip size. If is it too low it is slow. If it is too high 
you can skip too far into the good area where the slower reverse read 
speed becomes more noticeable. And skip more than half way to the next 
occurrence of the bad head and you could even miss data on the reverse pass.

First pass forward
111111111111111222222222222222333333333333333444444444444444
                                  SSS  SSSS SS  S SSS SSSS
Second pass forward
111111111111111222222222222222333333333333333444444444444444
                                  SSS SSS    SS  S SS  SSS
Second pass reverse
111111111111111222222222222222333333333333333444444444444444
                             SSS SSS   SS  S SS SSS

Hope this helps explain it. But you are still free to use my 
--no-reverse-pass options as you see fit.

Scott