[wellylug] Convert ReiserFS to Ext2

David Antliff dave.antliff at paradise.net.nz
Sun Apr 4 11:30:58 NZST 2004


On Sun, 4 Apr 2004, Enkidu wrote:
> It shouldn't be much slower overall. Firstly not every access is write
> and not every access update the journal, and secondly I believe that
> journal writes are very small and fast.

So do journal writes bypass the IO scheduler? If so, then they would
adversely affect the disk performance because the heads would need to
traverse back to the cylinders where the journal is stored (therefore
making journal writes small, but slow). If not, then the journal writes
will be priority-queued like any other disk write which would make the
whole point of journalling milestones completely pointless... unless of
course all this journalling logic happens below the disk access layer. But
the filesystem is at a 'higher' level than the IO scheduler so that
suggests journalling has little to do with the filesystem at all (apart
from having a journal, or not).

Since I'm not very familiar with the Linux disk i/o mechanism, I ask these
questions purely out of curiosity. Anyone got a link to a nice
how-it-works webpage?




More information about the wellylug mailing list