I think you have an interesting idea. Split the DB across disks so
concurrent activities are on different spindles, preferably at 15,000
rpm and use a multiprocessor/core cpu with the application threaded so
that you deploy more than one processor concurrently. If you could be
sure to schedule each thread to a different processor you could have a
poor mans processor cluster.
It would be neither "lite" nor simple, but could potentially be very fast.
Cesar David Rodas Maldonado wrote:
I don't know I just was wondering... but i think it will be faster if you
have a really great DB and when you and insert just will be locked the
table where you are doing the insert not the whole DB... what do you think?