Got my Terabyte!

Seems another excellent manner in which to permanently get rid of sensitive data is to send it to the Army records archive facility in St. Louis! They claim most of my active duty Army records were lost in a fire in the facility. Maybe I could send my old HDs there marked SECRET <G>

Best,

Dave
 
gismo said:
Chris, it's a generally accepted practice to implement improved error rates along with capacity, either by better magnetics or by using redundant (error correcting) data. I can't say for certain but I suspect that the 1 in 10^4 error rate is an average over the range of magnetic disk recording designs in use at the time this statistic was compiled and it's probably higher than what you'd expect from 1+ TB drives once they're beyond the infant mortality potential. Another factor is, assuming the error rate you quoted includes electronic and mechanical failures vs magnetic media failures) the non-media failures are not directly proportional to the capacity of the media.
Click to expand...

Yeah, when I run programs I just see errors, I don't get the luxury of knowing whether they were media induced, caused by electronics on the drive, caused by corruption on the SATA bus, or caused by errors in the RAM or CPU. I just see errors. The numbers I hear from folks in the hard drive industry is "somewhere between 10^14 and 10^16". Since all of these numbers are in the range of 1PB (about 10^16 bits), I find I have to apply end-to-end checksums on everything anyways... (and observe checksum failures multiple times per run)

The real issue is that with the larger drives people are tempted to use larger datasets (doing things like streaming TV shows, as I mentioned) and with larger datasets you up the probability of one of those errors actually hitting you. For TV shows, who cares (unless it causes your software to repeatedly crash). But I've just grown paranoid about losing data from my work... :-)

Chris
 
Dave Siciliano said:
Now that I have the new external HD; I'd like to dispose of the old one that has some bad sectors. Should I just erase the old HD before tossing it? Re-format. What to protect the old files from being used? I guess I could just copy old Gilligan's Island episodes on there (over write old files) but could never toss Mary Ann <g>

Best,

Dave
Click to expand...

We spike them here at the University. Sledge plus a railroad spike to put at least 3 holes through the whole thing.

Alternatively, if you have access to firearms, perhaps a trip to the range is called for. :D

But if you have the time, overwriting the data several times (I used to do 7 passes of junk) will make whoever is trying to get at your data have to spend far too much money to make it worth their while.
 
Skypilotcfii said:
Hammer in the Driveway and even DOD won't get any thing off it.
Click to expand...

Actually, wiping the disk stands a better chance of rendering the data unreadable to the most sophisticated attempts unless you shred the platters or dissolve the magnetic surface.
 
colohan said:
Yeah, when I run programs I just see errors, I don't get the luxury of knowing whether they were media induced, caused by electronics on the drive, caused by corruption on the SATA bus, or caused by errors in the RAM or CPU. I just see errors. The numbers I hear from folks in the hard drive industry is "somewhere between 10^14 and 10^16". Since all of these numbers are in the range of 1PB (about 10^16 bits), I find I have to apply end-to-end checksums on everything anyways... (and observe checksum failures multiple times per run)

The real issue is that with the larger drives people are tempted to use larger datasets (doing things like streaming TV shows, as I mentioned) and with larger datasets you up the probability of one of those errors actually hitting you. For TV shows, who cares (unless it causes your software to repeatedly crash). But I've just grown paranoid about losing data from my work... :-)

Chris
Click to expand...

I agree that all else being equal, increased dataset size will come with a proportional increase in errors but technology can improve things so and the relationship between errors and data quantity varies with the error cause. I can remember a time in the early days of solid state memory when the expected bit error rates were so high that single bit error correction was necessary for the memory sizes in use at the time and it was predicted that the next quadrupling of memory density would need multibit correction. But the major cause of bit errors in these memories was atomic decay in the integrated circuit packaging material (causing a high speed particle to alter the stored charge on a capacitive memory cell) and the scientists figured out how to make the silicon less sensitive to this and also improved the package materials in ways that reduced the potential errors. That was when the individual memory ICs had capacities in the hundreds of kilobits and read times in hte 100 ns range. If the BER of those chips were applied to today's multi gigabyte memory chips which have close to a million times higher bit density and more than 100 times faster there would be several failures per second but the BER rate has improved so much that error correcting memory is rarely used except for servers and mission critical systems. Granted there are differences between solid state dynamic memory and rotating magnetic storage but the same situation exists WRT the improvements in BER that come by design with improvements in storage density.
 
You must log in or register to reply here.
Back
Top