Anonymous (no@spam.com) on 9/11/08 wrote:
---------------------------
>Sigh, it is trivial to store enough energy in the flash unit to commit the buffers
>to flash on loss of power, which is what many HDDs do right now anyway, problems with that?
>


Can you point to any documentation for a hard drive that actually does this, sold in the past say 10 years? I have heard many times that hard drives supposedly can draw energy from the spinning disks to maintain power long enough to write cached data when power goes out, but I have several problems with it:

1) I don't believe drives can write at variable speeds, writing data to a decelerating drive doesn't sound possible to me, or at least not something that modern drives are capable of.

2) It would complicate the design of the drive for it to not only be able to power the spindle during operation, but also draw power from the decelerating spindle like a flywheel does when power is lost. Wouldn't a small capacitor and onboard flash equal in size to the RAM cache be cheaper, certainly for the past few years?

3) If this is being done, why is leaving the write cache disabled UNIVERSALLY recommended where data integrity is a concern, except in cases where you have an array with non-volatile (battery backed) cache? I would have expected those models that provided this facility to have a write cache that cannot be disabled. And sell for a premium price!

4) The story has the ring of an urban legend....i.e., I hear it mentioned here and there over the years, but have never seen it referred to by any drive manufacturer, storage white paper, etc.

I've done a lot of consulting in storage over the past 10 years (EMC, XP, that sort of thing) so I actually do read storage whitepapers from time to time, and have NEVER read anything about this except in forums like this one. If I was Seagate and I sold drives that could do this, I would not hide this ability like it was a state secret.