Emmanuele Bassi <[EMAIL PROTECTED]> writes: > On Sun, 2007-02-11 at 15:47 +0100, Ole Laursen wrote: >> I would have thought that flash memory and journaled file systems were >> a natural fit since you don't have the seek overhead that are present >> with disk-based memory. > > ext3 is a "heavy" file system, with writes occurring in bursts due to > the journal, even when filtered through the wear leveling mechanisms > that flash-based memories apply (hence the usual suggestion of mounting > flash memories using the async parameter, delegating the writes to the > card's controller paired with the kernel queue).
I'm sorry, but I still don't quite understand. :-( File access is bursty in nature AFAIK. Having to write to a journal/log increases the number of writes, and async can certainly help in making it possible to combine several log updates in one write. Is that what you're referring to? If it is, the question is how many extra written blocks it causes, and whether the overhead is worth the increased security. -- Ole Laursen http://people.iola.dk/olau/ _______________________________________________ gnome-devel-list mailing list [email protected] http://mail.gnome.org/mailman/listinfo/gnome-devel-list
