[Fwd: RE: Fop ALT-Design]

["Peter B. West" <pb...@powerup.com.au>]

Would those of you who are much better versed in these things please 
reply to the outstanding portions of Hans Kappert's letter, most of 
which is included below.  I have already advised him to subscribe to 
fop-user and I have apologised for my delay in replying.

Thanks.

Peter

-------- (part of) Original Message --------
Subject: RE: Fop ALT-Design
Date: Thu, 30 Jan 2003 10:00:41 +0100
From: Hans Kappert <h....@co-maker.nl>
To: 'Peter B. West' <pb...@powerup.com.au>



Peter,

Thank you for your quick reply. Of course, you may CC your reply to
fop-dev. Will they mail me or do I have to monitor a mailing list?

I will evaluate release 0.20.5.

Regards,

Hans

  > From: Peter B. West [mailto:pbwest@powerup.com.au]
  >
  > Hans,
  >
  > I would like to CC this reply to fop-dev, with a follow-up there,
  > because the questions you ask will be much better answered by
  > others on
  > the list.  It's true that I have been working in memory consumption
  > problems, but they are only in the front-end of the processing; the
  > generation of the FO tree.
  >
  > I am pleased to say that the other developers are now interested in
  > integrating my work into the mainline redesign effort (provided I get
  > the necessary notes together.)  Because of the areas in which I have
  > been working, I am not as familiar with the details of performance of
  > the existing system as many others, which is why I would like to pass
  > your queries on.
  >
  > This re-integration will have benefits for the next
  > generation of FOP,
  > but notable improvements have already been made, and I
  > suggest that you
  > evaluate the release candidate(s) and final release of 0.20.5.
  >
  > Please let me know if it is OK to post your message with this
  > reply to
  > fop-dev.
  >
  > Peter
  >
  > Hans Kappert wrote:
  > > Dear Peter,
  > >
  > > We are very glad to see that you are busy with improving
  > FOP 's memory
  > > consumption and size-independency.
  > > We now use fop-0.20.4 and see memory consumption upto 80Mb for
  > > converting a 800Kb XML file to a .pdf file (no forward
  > references used).
  > > It also took 15 -20 seconds on a 2Ghz pentium 4 with 256Mb RAM.
  > >
  > > We would like to use fop for reporting/printing purposes in an
  > > application we developed. This application will need to
  > print invoices
  > > in xml format leading to thousands of pages of PDF. What is
  > your advice
  > > on the folling issues?
  > >
  > > - Is it a good idea to split the XML file into smaller bits
  > and issue
  > > multiple FOP requests?
  > > - If so, how big is a good maximum for a server with 1GB
  > internal memory.
  > > - When the FOP handles two requests simultaneously (using
  > two threads),
  > > does the JVM need two times as much memory with the -Xmx
  > option (See
  > > http://xml.apache.org/fop/faq.html#faq-N10117)?
  > >
  > > - If so, can we prevent the servlet runner (Tomcat) from
  > starting more
  > > threads when requests keep coming in?
  > > - Is it a better idea to wait for ALT-Design to become stable?
  > > - If so, do you have an indication when? Or, if hard to tell, what
  > > year   ;-)
  > >
  > > Any response would be of great help..

-- 
Peter B. West  pbwest@powerup.com.au  http://www.powerup.com.au/~pbwest/
"Lord, to whom shall we go?"


---------------------------------------------------------------------
To unsubscribe, e-mail: fop-user-unsubscribe@xml.apache.org
For additional commands, e-mail: fop-user-help@xml.apache.org

Re: [Fwd: RE: Fop ALT-Design]

["J.Pietschmann" <j3...@yahoo.de>]

Peter B. West wrote:
>  > Hans Kappert wrote:
>  > > - Is it a good idea to split the XML file into smaller bits
>  > and issue
>  > > multiple FOP requests?

It depends.

>  > > - If so, how big is a good maximum for a server with 1GB
>  > internal memory.

I'd say some 900M but you can experiment: write a small Java program
which allocates a hunk of memory (using new byte[size] and zero it),
then null the pointer and start over with a larger allocation size,
using, say, 10MB steps. Use the system monitor to detect when excessive
swapping sets in. Dont forget to preload any other services you'll
expect to run and keep them busy. Add 50M or so as safety margin.

>  > > - When the FOP handles two requests simultaneously (using
>  > two threads),
>  > > does the JVM need two times as much memory with the -Xmx
>  > option
If you are past the allocation of 256M, then basically yes.

>  > > - If so, can we prevent the servlet runner (Tomcat) from
>  > starting more
>  > > threads when requests keep coming in?

Put the actual PDF rendering into a sync'd static routine of
your servlet or use a singleton, as usual. Look into any book
covering Java MT programming in somewhat more depth than the
usual "Java in three days".

J.Pietschmann


---------------------------------------------------------------------
To unsubscribe, e-mail: fop-user-unsubscribe@xml.apache.org
For additional commands, e-mail: fop-user-help@xml.apache.org