phk@phk.freebsd.dk wrote: 

> I gave up on journalling myself because IMO it complicates
> things a lot and the problem it solves is very very small.

If only hardware manufacturers were to equip hard drives with
a mechanism to ensure atomic writes. A capacitor large enough
to hold enough energy to flush the cache upon detecting the
power supply was cut would be sufficient. They could even use
a battery the status of which could be monitored via S.M.A.R.T.,
I don't see how implementing something like that could possibly
make the cost noticably higher.

Recent IBM Thinkpad and Apple PowerBook G4 laptops have sudden
motion sensors which park the disk heads when a sudden motion
is detected in order to prevent damage from a fall and similar,
so this atomic write guarantee mechanism should be trivial for
them to implement and it would save us a lot of work.

> The impact in disk seeks is non-trivial to predict, but it is
> very hard to argue that it will not lead to an increase in
> disk seeks.  (This is really a variant of the age old argument
> between jounaling filesystems and "traditional" filesystems)
>
> I can only recommend that you try :-)

Journaling would definitely cause more seeks, but the question
is how well that can be compensated for by the reduction of
writes gained from using the other speedup mechanisms I proposed.
I might try implementing this in GBDE first to see what the
benchmarks say.

ALeine
___________________________________________________________________
WebMail FREE http://mail.austrosearch.net