I wrote very similar code to that below and I noticed no appreciable memory
usage difference comparing (case1 to case2 to case3)
//case1
for(long i = 0; i < Long.MAX_VALUE; i++) {
logger.warn("The value of i is " + i);
}
//case2
for(long i = 0; i < Long.MAX_VALUE; i++) {}
This is what is so perplexing.
That code would also take on the order of an hour to execute (regardless of
log level)
while as Sean suggested:
//case3
for(long i = 0; i < Long.MAX_VALUE; i++)
{
if (log.isWarnEnabled())
{
logger.warn("The value of i is " + i);
}
}
would take only ~60 seconds.
The performance implications are clear....Memory is a different story
altogether and of course consume enough memory and you start to have even
worse performance!
-----Original Message-----
From: Ceki Gulcu [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, December 18, 2002 4:43 PM
To: Log4J Developers List
Subject: RE: Large Memory usage while using log4j
The following will create oodles of StringBuffers and Strings although
these will be eventually garbage collected.
Logger logger = Logger.getLogger("test");
for(long i = 0; i < Long.MAX_VALUE; i++) {
logger.warn("The value of i is " + i);
}
Do you know what types of objects are created?
At 16:23 18.12.2002 -0500, you wrote:
> That is true for our debug level logging, however we also have
> info level
>and warn levels that utilize string concatenation. We do understand we are
>paying for the string concatenation of the function call and thus creating
>several transient string objects that will be destroyed after the function
>call returns. This however should cause performance implications with a
>small memory cost. The concern is we are seeing a tremendous memory cost
>even though our log4j settings are configured to log nothing.
>
> We use this logging scheme because we have several levels of
> deployment.
>During development test cases are run with debug level, during Official
>Testing we run at a level of Info or Warn, and crank it up to debug if
>needed (by category), during Release we set our logging to error (which for
>us is the same as none). In general our ideology has been one of robustness
>and maintainability come before performance however an order of magnitude
>memory increase/decrease is something that captures our attention.
>
>Jim
>
>-----Original Message-----
>From: Sean Reilly [mailto:[EMAIL PROTECTED]]
>Sent: Wednesday, December 18, 2002 4:00 PM
>To: Log4J Developers List
>Subject: RE: Large Memory usage while using log4j
>
>
>Just to sanity check....
>
>if you log a message that performs string concatenation, you do it like
>this, right?
>
>(assumes an org.apache.log4j.Logger instance log)
>
>if (log.isDebugEnabled())
>{
> log.debug("Test Log Message: " + foo + " = " + bar + ".");
>}
>
>Sean
>
>Sean Reilly
>Programmer, Point2 Technologies, Inc.
>(306) 955-1855
>[EMAIL PROTECTED]
>
>
>
>-----Original Message-----
>From: James Gallogly [mailto:[EMAIL PROTECTED]]
>Sent: Wednesday, December 18, 2002 2:55 PM
>To: Log4J Developers List
>Subject: RE: Large Memory usage while using log4j
>
>
>
> Thanks!! That is a good thread, what concerns me the most however
> is that
>nothing is being logged.
>
> Our logging was set to error (we do no error level logging). This
> means
>that nothing is logged in either case.... does this mean that the
Conversion
>Pattern is applied before the decision to log or not occurs? In other words
>why is log4j using up so much memory to do what should be nothing?
>
>Thanks,
>Jim
>
>
>-----Original Message-----
>From: Ceki Gulcu [mailto:[EMAIL PROTECTED]]
>Sent: Wednesday, December 18, 2002 3:42 PM
>To: Log4J Developers List
>Subject: Re: Large Memory usage while using log4j
>
>
>
>The following thread might be relevant:
>
>http://marc.theaimsgroup.com/?l=log4j-dev&m=103366188809758&w=2
>
>
>
>At 15:35 18.12.2002 -0500, you wrote:
>
> > On my project we have been heavily utilizing log4j as our
logging
> > solution.
> >Recently we ran our code under very heavy load so as part of our
> >optimization we wrote code to extract the log4j logging calls from our
> >source. We did this in hopes of reducing some of the overhead of
> >concatenation strings together thus squeezing out some performance gain.
> >
> > Much to our surprise the greatest impact was in memory usage,
our
> > memory
> >usage went from 1 gig to 100 meg. This order of magnitude decrease in
> >resource usage caught our attention to say the least. How might we be
miss
> >using the log4j tool to cause such large memory consumption?
> >
> > We are performing small scale experiments trying to duplicate
the
> > essence
> >of the problem but have had little success so far. Any help or direction
> >will be much appreciated.
> >
> >Thanks,
> >Jim
>
>--
>Ceki
>
>
>
>--
>To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
>For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
>
>
>
>--
>To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
>For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
>
>
>--
>To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
>For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
>
>
>
>--
>To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
>For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
--
Ceki
--
To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
--
To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
