Failures depend on whether the flash is being written to regularly or not.
We built a Linux gateway box that used an SD card as the file system.
They would fail within about 2-3 months of operation. We used SandDisk
SD cards. SandDisk does not implement any wear algorithms, hence the
quick failure. So the lifetime is a matter of write once only. Also many flash
cards do not implement error correction or automatic re-mapping. So
there will be a wide variation in experiences and quality of Flash. The push
for high density and high speed has driven the market more than reliability.
I program embedded devices, and the flash memory on a Freescale
microprocessor is a lot more robust than today's flash cards, so there
is no reason why the technology cannot produce long life, just that the
market has not pushed that as a design feature. Whereas the Freescale
microprocessor has to be reliable, so it has a different performance metric.
But the super fast burst rate of a camera has speed as the dominant
profit margin maker for compact flash, not longevity.
Now I hear that the Akashic record is permanent, but that exists on the
astral plane and to date, no devices have been built to access it other
than well trained psychics.... :-) but I digress...
WayneS
At 11:37 AM 11/6/2008, Ken wrote:
>To date, Flash Memory has not been considered to be "archival" by any
>means. In fact, within the telecom industry it has already been proven to
>be, uh, less than stable. I understand and agree with what Wayne is saying,
>but in actual practice we're experiencing failures. From what I'm hearing
>from my cohorts on the supply side--it is looking like about 5% of 36 months
>of in-service devices are experiencing corrupted memory. Inotherwords, one
>out of every 20 boxes is getting random failures pointing directly to
>corrupted flash memory withing three years. This 5% failure rate definitely
>coincides with our own experiences. What really gets one sweating is when
>you have a device that needs rebooting and it doesn't come back online
>because the programming and configuration, which is stored on flash memory
>is corrupted. Trust me, it happens more often than anybody wants to admit.
>That's why, in our own company, we have a pretty tough procedure to follow
>before we ever cycle a box.
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|