[Bug-ddrescue] On speed

[andrew zajac]

Regarding the topic of speed of data recovery using GNU ddrescue, I use crappy USB tri-head (ide and sata) interfaces to recover data from failing drives.  I can usually get about 12 MB/s top speed and can image a drive in a few hours with that.  Although I seldom do.

For example, the drive I am currently doing is a 120 gig 2.5 drive from a Mac.  It stopped being detected and the computer would not boot.  It was replaced.  I got the failed drive to be recognized by patiently power cycling the drive (unplugging and plugging the power cable) until it powered up properly and was recognized.  It only needs to get there once after which you don't unplug it again.

I tried imaging the drive and got about 5 gigs over the first hour.  I had to reset it about a dozen times during this period (unplug and replug the usb interface; hotplugging).  It would shriek and click before it would conk out, a cue that I would pick up on from about thirty feet away.  So I would know to drop what I was doing and restart the recovery.

After a few passes, I decided to try the much slower direct mode, which has performed very well.  It has been going for almost 14 hours now without any more errors.  I reckon the high ratio of error (115 gigs of error out of a 120 gig drive) is due to the high speed since I am reading the parts of the drive that I wasn't able to read at regular speed (with no new errors). Although its rate is currently only 368 Kb/s and if this keeps up, it will take until Friday morning to complete, I don't care about the speed.  Or rather, there's nothing I can do about it. 

As long as the drive is spitting out data, it's a success.  Very few of my customers are under a time crunch that requires the job be done in such a short time anyway.

This phenomenon is not unique to this drive.  I find that most of the problematic drives I work on behave like this. 

Would I get better rates in direct mode using a more sophisticated interface?  I dunno.  But there are two show-stoppers that prevent me from exploring that possibility.

The first is cost.  I have fried a few components by plugging in my customer's faulty hardware.  Data recovery is usually "no data - no charge".  So there's no way that I am risking more than a few dollars worth of equipment on questionable hard drives, especially faced with the possibility that these cases are hopeless and therefore I won't get paid.

USB Tri-Head adapters cost less than ten dollars each, including the power supply.

The other problem is the eSATA interface.  On the machines on which I have tried an eSATA interface, the problem is that they are not hot-swappable like USB devices.  In one case, the drive is not recognized unless it is present at boot time.  Since I often need to power cycle the drive several times to get it to "come alive", I would need to reboot the computer each time I power cycle it to get the bios to check to see if it can recognize the drive (the motherboard needs to see it before the kernel can assign a /dev to it.)  Annoying.

In another case, the only way to reset the drive was to hit the power button, which would cause the drive to reboot and in many cases, once you get the drive to "come alive" you never want to power it down.  A USB interface can be plugged and unplugged without powering off the drive.

So, is there an sSata interface out there that can cycle the eSata interface without powering off and on the drive?

So, I'm all for high speed, but not at the cost of leaving data behind.

Cheers!

Andrew Zajac
AndrewZajac.ca
ubuntu-rescue-remix.org

---

Be smarter than spam. See how smart SpamGuard is at giving junk email the boot with the All-new Yahoo! Mail