[rescue] expanding a ZFS pool: leave slack space?
Romain Dolbeau
romain at dolbeau.org
Sun May 2 13:06:53 CDT 2021
Le dim. 2 mai 2021 C 19:20, Phil Stracchino <phils at caerllewys.net> a C)crit
:
> I hear the consistency of units argument. But the storage industry
> didn't switch to using strict SI units (...)
Genuine question: did they actually switch? Or did they almost always
used powers of ten? I'm not talking about the very early days when IBM
or DEC would have dedicated salespersons selling disks that could
barely reach millions of bytes directly to an institution's buyperson.
I'm talking when the storage products started to hit the shelves
to be bought by smaller companies or individuals, and there was
small actual "marketing" packaging to print something on.
I can't remember a time when I didn't hear the complaint that "it's
not a proper mega/giga/terabytes". But then my memory is not
counted in any of those units and might be going :-)
... answering a bit to myself after checking the web: they probably
did, as the Quantum Prodrive 40S had 82029 sectors of 512 bytes
or almost 42MB, and was sold as "40 MB" (it was really 40 MiB,
but binary prefixes didn't exist back then). On the other hand,
the Trailblazer 850 is documented as 1647c 16h 63s (of 512 bytes)
or almost exactly 850 MB - or 810.6 MiB.
So it happened a long time ago, and they were *never* ambiguous
on Tera and Giga, only on Mega - which the floppy was already using
to mean 1024*1000, so even then it was not clear-cut.
> And now we're in the bad place where gigabyte means a different thing
> depending whether you're talking about memory or storage.
That's where we disagree. Gigabyte means 10^9 bytes. The fact
memory vendors (and *only* memory vendors, no-one else does
that) can't be bothered to add a lowercase 'i' on their packaging
and website is the only source of confusion.
> And memory
> can't be simply declared to be powers of 10 instead of powers of 2 and
> still come out to manageable numbers.
That's an old argument which has never made sense to me. No-one
asked to change memory to power-of-ten or to say it's a 34.36 GB
DIMM. Just to put "32 GiB" instead of "32 GB" on the packaging, as
per the relevant IEC standard. That's it, problem solved.
It removes the (non-existent to anyone used to SI) confusion and
is a trivial change that could have been made over 2 decades ago.
> And then we get into flash memory devices that use storage
> industry gigabytes on top of a memory physical layer, and it's just a
> hellish mess and all capacity markings are approximations.
Flash is storage and sits behind a controller that does a lot
of magic to hide the complexity (block update, write levelling, ...),
and they don't ship in Po2 value anyway. Also there's usually a
lot of chips - some of which may not be present on some models
to create a range. There's absolutely no reason for flash or any
other form storage to not use decimal SI prefix. So they do.
The only reason not to obey the standard is to give a motive
to a handful of old-timers who enjoy a good flamewar on the subject
every now and then :-) It's been a long time, I guess the younger
generations don't care so much...
Cordially,
--
Romain Dolbeau
More information about the rescue
mailing list