Re: My problem is different

[Alon Lubin]

Sorry for the late response,
But I don't understand why this would help.

Instead of generating one simple statement that takes few seconds to complete,
The binary logs contains GB's of unneeded information in the format of RBR.

Transaction division will not speed up the IO required for this process, the slaves still need to read thousands of rows from the binary log.

Now, what really interests me is why not use SBR when working with "read-committed" isolation level?
Why is it different than "repeatable-read" isolation level in which SBR are enabled?

Both don't force locking, allowing change of underlying data during a transaction. The only difference is that repeatable-read will always show the same information when performing query, while read-committed shows available data only.

Thanks, Alon.