Jay A. Kreibich wrote:
> Yes. Hence the "and this is the important part" comment. Most of
> the time when people are building billion-row files, they're building
> a new DB by importing a static source of data. If things go wrong,
> you just throw out the database and try again.
That's kinda like doing it all in a single big transaction, which I
wanted to avoid. :)
But my take-away from this is conversation is generally "you're SOL":
we are backed by a tree, so we have to update existing nodes to point
to the new ones, so we have to touch existing pages on INSERT. Is that
about right?
It's not in any way a result of my schema? My primary key is a pair of
integers A,B. The first column in this particular use case is in the
range A = [0, 2million) and the second is in the range B = [0, infinity).
We
insert records A=0->2million, B=0, then
insert records A=0->2million, B=1,
etc.
Could this have an impact on how many pages need to be touched on
INSERT?
> It would also help to bump the cache up...
That works great until the db size blows through the total RAM on the
system, at which point we're of course disk-bound again. At the moment
I'm only inserting about 4k rows/second. :/
Eric
--
Eric A. Smith
I have always wished that my computer would be as easy to use as
my telephone. My wish has come true. I no longer know how to use
my telephone.
-- Bjarne Stroustrup
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
