NetBSD-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: state or future of LFS?

>>>>> "mf" == Martin Fouts <> writes:

    mf> most consumer NAND flash devices such as usb sticks,
    mf> tend to have a microcontroller that hides block rewriting and
    mf> wear leveling, so you don't really have access to the raw NAND
    mf> anyway.  I believe that the current generation of SSDs work
    mf> that way, but I'm not familiar with them.

yeah, they do: They are mostly SATA.  The most desired ones of which
I've heard are the Intel X25E and the STEC Zeus which both have
controllers you can't bypass.  Most people buzzing about them think
proprietary software inside the controller is the secret sauce that
has suddenly made hte expensive SSD's buzzworthy.  Then there is a
separate class of inexpensive SSD's from $shadyvendor on newegg that
are not buzzworthy, are slow and just for making cheap laptops.  It is
mostly but not just the hidden filesystem---also I think they
sometimes may have weird things inside them like supercaps and RAM
buffers.  In any case, reviewers have found the Intel models, which
remain well-respected, get ``fragmented'' and perform like a quarter
as well as they do when unfragmented:

so they must have their own proprietary software LFS-like thing inside
the device, which has good and bad features but FWIW is mostly what
they are selling you.  Who knows, maybe it is even the one you wrote!

there are some which are PCIe rather than SATA, but there's not much
experience with them that I've heard, and I think they still have
controllers which can't be bypassed:

among USB sticks, I've heard FUD that the wear-leveling soemtimes
works in 16MB chunks, so if you have an data-overwrite filesystem with
a metadata journal, you can blow the chunk that contains the journal.

Based on that, it might be worth doing wear leveling even above the
controller, but you can't do much about NAND read-disturb.  Also when
you yank USB sticks without detaching first, they can corrupt data far
away from anything to which you were writing:

I think there was some statement from Crucial that they do not have
this problem, but the cost of the simple way they managed to not
having the problem is that the wear leveling isn't as good---I don't
remember the details and can't find the PDF right now.

    mf> Since I can't get the company that bought Danger to meet the
    mf> commitment I had made on Danger's behalf to release much of
    mf> what we had done under the BSD license, I have decided to
    mf> design and implement a NAND flash file system based on what
    mf> I've learned over the last decade.

and I suppose you will write it using the BSD license that has served
you so well in the past. :)


Attachment: pgpR55xOL4eS2.pgp
Description: PGP signature

Home | Main Index | Thread Index | Old Index