> > > Actually, we have postponed journaling to be the last thing to > > > implement - we will have a fully working BFS without journaling > > > first. > > > We are still planning how to do the journal part, but we also > > > plan > > > to > > > circumvent some shortcomings of BFS there (serialized log > > > access). > > > > One journal per allocation group. That would rock. :) > > Well, one of my disks has 500 allocation groups - I don't think that > this approach would be too wise... ;-)) Well... But that would be a way to do that... It would be esier to manage than other possible solutions and only access to the log by threads writting to the same allocation group would be serialized. Of course this is only one possibility... We can come up with something else... I gues concurrent access to *ONE* log would be kinda difficult to implement. -Bruno -- Fortune Cookie Says: If you stand on your head, you will get footprints in your hair.