On Jan 14, 2008 3:15 PM, Monty J. Harder [email protected] wrote:
On Jan 14, 2008 11:09 AM, Oren Beck [email protected] wrote:
Accumulate write counts to a buffer in session ram- and write to flash only as part of unmount, or shutdown where unmount was not explicit but presumed.
That shouldn't be necessary. Internally, flash drives save information along with each data block that isn't presented via the USBMSD interface. When I write Sector X of a filesystem to the flash drive, it doesn't always write the same physical location. Instead, it writes the new version of that sector to another location, and has some way of knowing which version is current. I don't know the internals, but it's probably using some kind of serial number that increments with each write.
At least that's how I would implement something like that. The device should be able to non-destructively determine how many sector writes it has performed, and how many sectors are nearing the expected lifetime.
That was part of my original concept- that such data was "present" merely needing to be accessed and externally readable.
The open queries are 2.
1. IS a device with 80% of it's estimated "life" used still "Stable" for archive use? Defined as whatever data is the "Last Writes" remaining stable for the same # of years as a written to "less" device or not. With qualifiers on confidence levels and non-obvious issues.
2. Agreement on flag or bit to declare a flash device "Archive" .
On Jan 14, 2008 9:16 PM, Oren Beck [email protected] wrote:
- IS a device with 80% of it's estimated "life" used still "Stable" for
archive use? Defined as whatever data is the "Last Writes" remaining stable for the same # of years as a written to "less" device or not. With qualifiers on confidence levels and non-obvious issues.
- Agreement on flag or bit to declare a flash device "Archive" .
AFAIK the device can be read as often as desired with no degradation. Thus, anything less than 99.999% life used up should be fine. Of course the lifetimes are estimates of averages anyway...
Jon.
On Jan 14, 2008 9:16 PM, Oren Beck [email protected] wrote:
- IS a device with 80% of it's estimated "life" used still "Stable" for
archive use? Defined as whatever data is the "Last Writes" remaining stable for the same # of years as a written to "less" device or not. With qualifiers on confidence levels and non-obvious issues.
Media is cheap. Media goes bad. I have yet to meet a media that I could reliable set in a fire-proof safe and depend on for true archival purposes, because either the media will go bad or the device interface will become obsolete. Even printing hard-copies with many modern InkJets is worthless, because the ink is only expected to remain readable for a relatively short period of time (even special inks are only expected to have a 100 year shelf life, and I maintain paper records older than that). Also, flash drives do have exposed metal contacts which can corrode or wear off.
Since media is continually becoming cheaper per MB, it only makes sense to "upgrade" your archived data every couple years. This allows you to consolidate stuff when it becomes scattered (it always does), and ensure that all of your data is still readable. I usually find that by doing this, the amount of physical storage room I need to store my data stays roughly the same over time, because larger capacity media becomes available. You should also make a habit of storing copies of applications needed to open your files.
~Bradley
-
Bradley Hook -
Jon Pruente -
Oren Beck