[phpa] Re: Congratulations

Viva,

Nick Lindridge wrote:
> > I was brought to attention to it by Tom Duffey because it uses Metabase
> > database abstraction package and it seems it had some problems with
> > if(defined()) { define(); ...} statements. I hope you can fix that soon.
> 
> This has been fixed for some weeks in release 1.2dev. On this note, and to my
> suprise, one user reported that they even Zend Cache (and phpa 1.1) broke
> their code but that release 1.2 is ok in this respect.

Ok, when 1.3 will be released so I can test it?


> > Anyway, it seems that my server is dying once in a while because my ISP
> > kills tasks that take too much CPU for a long while. Some pages really
> > require a lot of code in there. So, I think a little compiler cache
> > could help.
> 
> I'd be suprised that this is what's happening. Have they said that they do

Yes. It is a virtual server that has 150 clients. Each client has its
own Apache copy and mine is being killed too often.


> this? PHP will terminate requests that take more than a certain amount of

No, it is Apache that is dying. Maybe that is because I am still using
PHP 4.0.0 and it may be dying for unrelated reasons.


> time. I think 30 seconds is the default.  A cache will help as it'll increase
> throughput on the server and not hit the disk or kernel cache, but you
> probably have other bottlenecks too.

I use very complex classes and that may be the problem that makes Apache
processes take too much CPU.



> > However I have some concerns regarding stability of your cache engine.
> > Once you fix the problem mentioned above I will give it some tests.
> 
> If you're running linux then grab 1.2dev and see how it goes. The problems

What about under Solaris? Have you made any tests?


> with metabase also have nothing to do with stability. They were merely bugs
> that meant that a few specific code patterns would fail to be handled
> correctly.

Yes, I know. Either it worked or not.

 
> With this or other extensions, I'd try advise trying it and monitoring
> carefully in as realistic a test environment as possible that it
> performs as you expect until you are confident.

Yes, I have a little time so I was hoping to learn from experiences of
others, concretely real world benchmarks. Do you have way to collect and
process statistics like with Zend Cache?


> > Another thing is that I have Zend Optimizer enabled. I know that is I
> > use your cache engine it won't use Zend Optimizer. My question, is does
> > it compensate to use the cache engine without the optimizer? Do you have
> > any benchmarks?
> 
> Benchmarks generally suggest that the cache improves performance above that
> attainable with the optimiser. PHPA also includes a code optimiser that
> can be enabled by setting phpa.tweaks = on. However in the releases this
> isn't developed as much as it could be and doesn't make a huge difference.
> Other tests in development make rather more of an improvement, but these
> won't get added for some time.

Good! So, you cache the optimized code, or you have to compile first and
optimized every time before you run?


> > Finally, I wonder what you think about developing and Zend Encoder
> > competitor. Is it feasable for you?
> 
> I think that there's already an opensource one. If I do anything next it'll

Do you have a URL?


> mostly likely be incorporating some more aggressive optimisations that I've
> played with. I'd like to increase execution speed further now that the loading
> performance has been reduced.

Great!

Keep up the good work,
Manuel Lemos
------------------------------------------------------------------------
  www.php-accelerator.co.uk           Home of the free PHP Accelerator

To post, send email to phpa@xxxxxxxxxxxxx
To unsubscribe, email phpa-request@xxxxxxxxxxxxx with subject unsubscribe

Other related posts: