On 1/19/07, Nicolas Williams <[EMAIL PROTECTED]> wrote:
On Fri, Jan 19, 2007 at 12:28:55AM +0100, Nemanja Corlija wrote:
> Even though this is a once a week thing, I still need to speed it up.
> I see 2 possible solutions:
> 1. Find some way to move file db to memory db and back at near disk
> speed (avoid creating index every time)
Why don't you use a trigger to duplicate INSERTs into the memory db to
the disk db?
Nicolas, thanks for the idea. Didn't really think of that.
But in my experience inserting into disk db with a unique index is
very slow when you have millions of rows. I think last time I did that
table had around 6-7 million rows and it took 2 hours or maybe even
more to insert 500.000 rows.
So its actually faster to load db into memory and create index there
two times, then it is to use a file db all time to avoid recreating
index on every use.
Thanks
--
Nemanja Corlija <[EMAIL PROTECTED]>
-----------------------------------------------------------------------------
To unsubscribe, send email to [EMAIL PROTECTED]
-----------------------------------------------------------------------------
