On Thu, 2006-08-24 at 10:54 +0100, Ted.Harding@nessie.mcc.ac.uk wrote:
I was also wondering about how long data would "stay in place" over time anyway. Does the "bit-state" of an element on a flash module tend to "diffuse" away over time?
I am not sure what the data persistence is of Flash memory but I would say that it should at the very least be better than a hard drive. The maximum write thing is improving all the time, which in combination with a wear levelling algorithm should mean that flash has the potential to be more reliable.
The one problem is that at the moment (excepting single bit failures) flash memory failures tend to be quite unpredictable. As Barry is currently finding out many (if not most) hard drive failures are predictable if you are monitoring the health status of the drive. This is because most hard drive failures are mechanical which have a tendency to degrade into a failure condition fairly slowly.
What I fully expect to see very soon are hybrid devices with enough on board intelligence to move data between flash and disk based storage based on access patterns. This is better than the vista approach because it is not using bus bandwidth. It will also (at least initially) be more accessible as I think it will be a long time before Flash can hit the cost per megabyte at the storage densities we are now used to.
We've been working with Microsoft on ways to speed up booting and application loading by complementing the hard drive with Flash."
This is already in the Vista betas, It can use a flash key drive as a sort of swap space/disk buffer. It's all done in a pretty clever way where unplugging the flash disk won't break anything and it evaluates each keydrive to determine whether it is fast enough to actually provide a benefit.