On Fri, Jul 24, 2009 at 06:28:58PM -0700, Robert Hodges wrote: > Here¹s another question for the team: would it be possible for the > replication plug-in to turn around and store what it gets in a table? This > would give you transactional storage of data in some transport-ready form > for an external daemon that then grabs it from the table. If this is > possible would it be transactional, in other words how could it be coupled > with the existing transaction so you would get correct recovery? > > I ask this because it always struck me that MySQL replication used a > less-than-optimal approach of storing things in binlog files when there was > a nice database standing by.
This is pretty much the vision I have (possibly straight out stolen from Brian) for having it work. The optimisation that could be done by engines would be to not store the repl event but rather just pointers to the data it needs. One big advantage is write speed. in bursts of writes, we're only writing linearly to the engine REDO log. -- Stewart Smith _______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
