throughput

[David Relson]

At 10:20 AM 1/26/03, Greg Louis wrote:

>On 20030126 (Sun) at 0820:52 -0500, Greg Louis wrote:
>
> > Either there's a db scaleability problem, or some changes (better
> > locking?) done since 0.8.0 have degraded throughput rather
> > significantly.  Now mind you, I'm on db-3.1.17; should I try upgrading
> > to 4, do you think?
>
>Tried that.  No difference in the time to process test20 starting cold
>(the spamlist did build rather faster, though; 11 minutes instead of
>16).

I'm lost track of the details.  How about a summary, i.e.

.MSG_COUNT = xxxx

         db-3.1.17       db-4
0.8.0   time1           time2
0.10    time3           time4

with as many of the 4 numbers as you have.

Can you hack the code to disable the new locking to measure _that_ effect?