Ole Ersoy a écrit :
I'm going to read this email as soon as the smoke
clears...
Nothing really complicated :)
The idea is to do a performance measure before doing some optimisation
on some part of the code, so Alex will have a clear vision of what kind
of performance improvment he will obtain.
Here, what we can say from those tests, is that the indices are not
really optimal : it seems that the time to grab entries grow linearly
with the growth of the database. For instance, from 300 000 entries to
400 000 entries, we have an increase in time to search for some entries
which is similar than from 200 000 to 300 000 : 6.04 sec for 200K, 7.14
for 300K, 9.24 for 400K. We would excpect the growth to be logarithmic.
Alex is working on that currently.
There is a page on confluence where we have put some thought about
backend, and it would be cool to add any comment, thought, improvment,
ideas :
http://docs.safehaus.org/display/APACHEDS/Backend
Emmanuel.
