DBTNG provides transaction support. We'll see a big improvement on insert performance if we wrap our inserts into a single transaction, as much as possible. Particularly important for d.o, since we're going to be doing probably in the tens of millions of inserts when we build the data.
Comments
Comment #1
sdboyer commentedWhoops, wrong queue. We probably could earn a fair bit from this, but hopefully we won't NEED to experiment with it until sprint 9.
Comment #2
sdboyer commentedPotential performance issue, retagging to sprint 10. Otherwise will leave it for phase 3.
Comment #3
sdboyer commentedGonna be revisiting this next week when we finally start calculating just how much volume util can handle.
Comment #4
eliza411 commentedComment #5
marvil07 commentedSeems like we have enough hardware for now :-p, but yes we need to do this, but probably after generic-parser-framework branch merge back to mainline.
Comment #6
marvil07 commentedDuring triage.