Re: [PHP4BETA] PHP4b3 RC3 Memory usage
| From: | Andi Gutmans | Date: | Tue, 16 Nov 1999 15:55:47 +0000 |
| Subject: | Re: [PHP4BETA] PHP4b3 RC3 Memory usage | ||
| References: | 1 | Groups: | php.version4 |
| Request: | Send a blank email to php-version4+get-6293@lists.php.net to get a copy of this message | ||
Can you please try and use regular MySQL links and see if this makes a
difference?
(i.e. the process doesn't grow within the first 10k hits to an eventual
final size)
Thanks,
Andi
On Tue, 16 Nov 1999, Chris Chabot wrote:
> I am using persistent connections to MySql ... in both php3 and 4 though ..
> so that shouldnt realy make the difference unless the implimentation of
> mysql_pconnect changed.. :) also ill try to cc these messages to php4beta
> and phpdev :) ill have to subscribe to the dev list first though, which i
> hadnt yet cause of its high trafic :)
>
> -- Chris
>
>
>
> ----- Original Message -----
> From: Andi Gutmans <andi@zend.com>
> To: Chris Chabot <chabotc@reviewboard.com>
> Sent: Monday, November 15, 1999 18:56
> Subject: Re: [PHP4BETA] PHP4b3 RC3 Memory usage
>
>
> > Weird. By the way you should probably have these emails also go to
> > php4beta nd php-dev. Maybe someone there has encoutnered similar
> > behaviour. Anychance you can find what is causing this growth? Maybe you
> > are using persistent connections?
> >
> > Andi
> >
> >
> >
> > On Mon, 15 Nov 1999, Chris Chabot wrote:
> >
> > > Well yes, im prety sure its actualy bigger in memory usage :)) very
> simple way
> > > to prove it, on php3 i can run 90 processes without running out of
> memory, and
> > > with php4b3 RC-n i can only do around 45 till it starts hitting swap
> like crazy
> > > ...
> > >
> > > the shared part of the memory is about 1300 Kbytes .... of 7000 total on
> my
> > > bigest scripts ... that leaves 5700Kbytes of memory consumtion :))
> > >
> > > After disabling the growth is still there yes ... see previsous email
> which
> > > included all the posible test combinations .... the stabilisation just
> doesnt
> > > happen ... it bounces between 5000k and 7000k ... and performing a lot
> slower :)
> > >
> > > -- Chris
> > >
> > >
> > > Andi Gutmans wrote:
> > >
> > > > 16 bytes is probably pretty much the minimum the "maximum cache block
> size"
> > > > should be tuned to, because the main Zend value container (zval) is
> that
> > > > size and it is most frequently allocated/deallocated.
> > > > By the way, are you sure that PHP is actually taking more memory? When
> the
> > > > process forks then most of that memory is copy-on-write and isn't
> really
> > > > allocated. Only what is being used by the program is allocated. I
> can't
> > > > think of what has really changed memory consumption wise except maybe
> > > > improved in PHP 4.0.
> > > > After disabling the memory cache do you still see the process growing
> > > > during the first few K of requests and stabalizing?
> > > >
> > > > Andi
> > > >
> > > > At 11:35 PM 11/13/99 +0100, Chris Chabot wrote:
> > > > >As you could read in my previous email, it does make a 100 to 200k
> > > > >difference, but at a very large expense of speed on smaller scripts
> (on my
> > > > >box the difference between 50 requests/second and 400 requests a
> second..)
> > > > >So ofcource an option is always a good thing, but would you want
> people to
> > > > >switch that on ? :))
> > > > >
> > > > >---------------------------------------------
> > > > > Chris Chabot
> > > > > CTO, The Reviewboard Magazine
> > > > > e-mail: chabotc@reviewboard.com
> > > > >---------------------------------------------
> > > > >
> > > > >----- Original Message -----
> > > > >From: Andi Gutmans <andi@zend.com>
> > > > >To: Rasmus Lerdorf <rasmus@lerdorf.on.ca>
> > > > >Cc: Chris Chabot <chabotc@reviewboard.com>; Sascha Schumann
> > > > ><sascha@schumann.cx>; <php4beta@lists.php.net>
> > > > >Sent: Saturday, November 13, 1999 18:01
> > > > >Subject: Re: [PHP4BETA] PHP4b3 RC3 Memory usage
> > > > >
> > > > >
> > > > > > At 12:03 AM 11/11/99 -0500, Rasmus Lerdorf wrote:
> > > > > > >If these prove to make a significant difference in his case, we
> should
> > > > > > >probably make them compile-time configurable options and document
> the
> > > > > > >performance/memory tradeoffs.
> > > > > >
> > > > > > Yeah but I don't think it makes more than max. 100-200kb of a
> difference.
> > > > >I
> > > > > > just want to make sure.
> > > > > >
> > > > > > Andi
> > > > > > ---
> > > > > > Andi Gutmans <andi@zend.com>
> > > > > > http://www.zend.com/
> > > >
> > > > ---
> > > > Andi Gutmans <andi@zend.com>
> > > > http://www.zend.com/
> > >
>