Hi,
I'm currently optimizing/analyzing a Zope server and I noticed the
following:
The number of references to specific objects (in this case specifically
Squishdot comments) increases almost linear with the number of threads.
Does this mean each thread keeps it's own cache pool of objects? Do th
On Tuesday 21 January 2003 4:14 pm, Ivo van der Wijk wrote:
> Does this mean each thread keeps it's own cache pool of objects?
yes
> Do these
> references point to the same object instances, or are these copied
> as well?
They are independant copies
> I have the impression the memoryconsumptio
On Tue, Jan 21, 2003 at 04:27:05PM +, Toby Dickenson wrote:
> On Tuesday 21 January 2003 4:14 pm, Ivo van der Wijk wrote:
>
> > Does this mean each thread keeps it's own cache pool of objects?
>
> yes
>
> > Do these
> > references point to the same object instances, or are these copied
> > a
On Wednesday 22 January 2003 11:43 am, Ivo van der Wijk wrote:
> > > Do these
> > > references point to the same object instances, or are these copied
> > > as well?
> >
> > They are independant copies
> Why exactly is this? concurrency issues?
Yes.
> Is there a
> way to optimize this?
1. Obvi
Toby Dickenson wrote:
On Wednesday 22 January 2003 11:43 am, Ivo van der Wijk wrote:
Do these
references point to the same object instances, or are these copied
as well?
They are independant copies
Is there a
way to optimize this?
1. Obviously, use fewer threads. Do you have a component
On Wed, Jan 22, 2003 at 12:14:27PM +, seb bacon wrote:
> Toby Dickenson wrote:
> > On Wednesday 22 January 2003 11:43 am, Ivo van der Wijk wrote:
> >
> Do these
> references point to the same object instances, or are these copied
> as well?
> >>>
> >>>They are independant copies
>
On Wed, Jan 22, 2003 at 11:57:20AM +, Toby Dickenson wrote:
>
> > Is there a
> > way to optimize this?
>
> 1. Obviously, use fewer threads. Do you have a component that otherwise
> benefits from many threads?
>
No, except that the server is pretty heavy loaded sometimes, and multiple
thr
Ivo van der Wijk wrote:
On Wed, Jan 22, 2003 at 12:14:27PM +, seb bacon wrote:
Toby Dickenson wrote:
I'm working on a 2.3.x instance which I have already upgraded to 2.4.4.
It soudns asif a further upgrde is reasonable. Also, the 'activity' tab
is new to me (looks interesting ;), and
On Wednesday 22 January 2003 3:01 pm, Ivo van der Wijk wrote:
> It soudns asif a further upgrde is reasonable.
I think you will have much more joy with 2.6
> Also, the 'activity' tab
> is new to me (looks interesting ;), and caching settins seem to be ignored
> mostly until 2.6, right?
it work
No, the global interpreter lock will allow only one thread to run at any
time. Depending on your situation (when your Zope is cpu bound)
*decreasing* the number of threads may improve performance. Also make sure
you tune the checkinterval (pass the -i parameter to z2.py).
Stefan
--On Mittwoch
On Wednesday 22 January 2003 3:10 pm, Ivo van der Wijk wrote:
> On Wed, Jan 22, 2003 at 11:57:20AM +, Toby Dickenson wrote:
> >
> > 1. Obviously, use fewer threads. Do you have a component that otherwise
> > benefits from many threads?
>
> No, except that the server is pretty heavy loaded somet
Stefan H. Holek wrote:
No, the global interpreter lock will allow only one thread to run at any
time. Depending on your situation (when your Zope is cpu bound)
*decreasing* the number of threads may improve performance. Also make
sure you tune the checkinterval (pass the -i parameter to z2.py
On Wed, Jan 22, 2003 at 07:49:11PM +, seb bacon wrote:
> Hmm, I thought the GIL pertains only to SMP?
The GIL pertains to the python interpreter on any hardware.
--
Paul Winkler
http://www.slinkp.com
Look! Up in the sky! It's PATHETIC LOOKING NOSE MASTICATOR!
(courtesy of isometric.spacenin
Ivo van der Wijk wrote at 2003-1-22 16:10 +0100:
> On Wed, Jan 22, 2003 at 11:57:20AM +, Toby Dickenson wrote:
> >
> > > Is there a
> > > way to optimize this?
> >
> > 1. Obviously, use fewer threads. Do you have a component that otherwise
> > benefits from many threads?
> >
>
14 matches
Mail list logo
