"Alexander A. Blashko" wrote:
>
> [EMAIL PROTECTED] wrote:
> >I'd like to define a few terms here - my word usages
> >may not match
> >other people's:
> >
> >1. Logging - capturing and recording data about events happening
> >in the system.
> >
> >2. Journalling - capturing database transactions
> >
> >3. Mirroring - exactly replicating database activity in another location.
> >
> >Each of these techniques has pros and cons. I consider mirroring to be
> >the
> >best way to guarantee data integrity and possibly the lowest overhead,
> >since
> >in theory, all that's require is to clone the write, update and delete
> >operations to an alternate destination. Practice is often less simple, of
> >course.
> We'd like to define database mirroring by using log-file.
> we developed replication procedure by using triggers,
> that store update operations
> ( insert,update,delete ) in log-table.
> What do you think about it?
>
> Best regards,
>
> Alex Blashko.
>
> ************
I think of mirroring as a nonintelligent operation, like RAID. The ultimate
test, though, is whether anyone would notice which of the two database
copies were being used.
What you're proposing is something like what I'd call
"transaction cloning", except that it sounds like you're using the DBMS's
external processes instead of having the work being done by internal processes
invisible to the user (?)
regards,
Tim Holloway
************
