Re: G5 ECC (was Re: X-serve verses G5's)

In article <020220061251334552%NoSpamDammit@xxxxxxxxxxx>,
Mark Conrad <NoSpamDammit@xxxxxxxxxxx> wrote:
What amuses me is that the general Mac user population embraces such
concepts as "journaling" with no problem whatever, yet dismisses ECC
ram as something that only a paranoid Mac user would consider using.<g>

They never stop to consider that some, if not all, of the file system
problems that journaling helps to recover from are directly caused by
single bit ram failures.

To cause those kind of filesystem problems through memory errors, you
need to get a memory error on the in-memory copy of file system data
structures. These aren't big. Let's be generous, and say that there
are a couple meg of such data.

It would take on average several years for a single bit RAM failure to
hit that, at current RAM error rates.

But way...it's not even that often, because most of the in-memory copies
of file system data structures are not purposefully changed, and so
won't be written back out. An error in one of those might confuse the
system, but won't corrupt the disk. We need to get an error between the
time the system writes the data to RAM, and the time it gets written to
the disk. Basically, the error has to hit during the time the data is
in the file system's write cache.

That won't be very long. That will reduce the risk of RAM bit errors
corrupting the file system to one in several decades.

The file system problems that journalling helps with are caused by
system crashes or power failures that cause partial writes of file
system data structures. Many file system metadata operations require
more than one sector to be written for the operation to succeed, so if
something interrupts this, you can end up with an indeterminate state.
That's what the journal deals with.

--
--Tim Smith
.

Relevant Pages

  • Journaling UFS with gjournal.
    ... The lack of journaled file system in FreeBSD was a tendon of achilles ... We do have many file systems, but none with journaling: ... If one provider is given, it will be used for both - ... for gjournal (especially when your disk driver does support BIO_FLUSH). ...
    (freebsd-current)
  • Re: Journaling UFS with gjournal.
    ... The lack of journaled file system in FreeBSD was a tendon of achilles ... We do have many file systems, but none with journaling: ... If one provider is given, it will be used for both - ... for gjournal (especially when your disk driver does support BIO_FLUSH). ...
    (freebsd-current)
  • Re: SCSI module eata no longer loading automatically from initrd on Sid on i386
    ... you wanted an initial RAM file system with only what is required ... showed lots of modules but *not* eata. ... need to be in the initial RAM filesystem or not depends on when ... be explicitly loaded *before* the permanent root file system is mounted. ...
    (Debian-User)
  • Re: SCSI module eata no longer loading automatically from initrd on Sid on i386
    ... you wanted an initial RAM file system with only what is required ... There are four ways I know of for a kernel module to get loaded during boot: ... before the permanent root filesystem is mounted, ... be explicitly loaded *before* the permanent root file system is mounted.. ...
    (Debian-User)
  • OT/drift: when is a RAMdisk an appropriate solution
    ... include the "ram disk" component in your project. ... Sometimes, a physical RAM ... only to wind up going directly to regular old ordinary memory, ... testing the file system software. ...
    (comp.lang.c)