Daniel Ockeloen wrote:
On May 12, 2006, at 11:20 AM, Ernst Bunders wrote:
Daniel Ockeloen wrote:
On May 12, 2006, at 10:56 AM, Ernst Bunders wrote:
Kees Jongenburger wrote:
-Hybernate is better in performance but less flexible in it's
datamodel
approach.
it's the opposite
you mean it's performace is worse but it is more flexible?????
Does anyone really know ? has anyone made say a page that collects
a few items from a form
and puts it in the cms ?. If someone has these tests why have they
not shared them ? If mmbase
is 10x slower using tags than something else with the same ease of
use we need to take hard look
at it.
Well i think there is a long standing issue with commit times in
mmbase. under certain curcomstances (I'm not sure what they are, but
i have many times witnessed the phenomena in the wild) mmbase node
commit times can exceed database commit times by factor 10. So that
is a real performance issue. I think the cache release strategy
(version 0.1 as they are) do a good job boosting performance (and
also reducing db load), but they still have to prove themselves under
various circumstances.
Thats why michiel and i want to add jmeter scripts to the releases.
That may be a good dear, but you need a little more i guess. you also
need a standardized application and dataset to test on. Also creating
a test that realistically mimics what happens inside mmbase on a
bussy day is not so easy. Perhaps it is better to add more live
performance monitoring tools, so you can see what happens in the real
world.
Another thing on this account is adding the ability to detect the
absence of important db indexes, or other problems that keep mmbase
from running well, and at least notify the adiministrator. but i
think extensive live monitoring possibilities would be quite the
thing to find out how well mmbase is doing. I added quite detailed
performance monitoring to the cache release strategies for that purpose.
Ernst
Well if we don't have proof that we can use and check what is wrong we
will not get anywhere ... I fully agree with you Ernst but if we never
start to compare but keep it claims 'mmbase is slower than x when doing
y' it just stays a claim. Mistakes are made there are several in 1.8.0
that effect performance but its untested so we didn't even notice in
all the nightly builds (i yesterday fixed one that resulted in a 10x
speedup on some things).
I do agree with you. It is definitely a good idear to have a
standerdized performance test.
Daniel.
_______________________________________________
Developers mailing list
[email protected]
http://lists.mmbase.org/mailman/listinfo/developers
_______________________________________________
Developers mailing list
[email protected]
http://lists.mmbase.org/mailman/listinfo/developers
