cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Gerhard Froehlich" <g-froehl...@gmx.de>
Subject AW: IT'S TIME TO REDO YOUR TESTS, GENTLEMEN!
Date Wed, 22 Aug 2001 20:57:11 GMT
Hi,
just running another load tests against cocoon with the latest
sources. I set up the testing like berin recommended in his posting.
I didn't get memory issues anymore. memory consumption was always
a little bit higher then initial heap-size and constant.

The only thing I worry is the cpu load on my machine (AMD 1,2GHZ). Only
5 simultaneous threads are enough to eat 90% percent cpu time on my machine.
I tried to trace down the problem (but this are really suppositions and
they're superficial. I have no time to get deeper this days. In the autumn I
will
take a longer rest from paid-work, only for cocoon of course :-)):

According to my tracing tool (OptimizeIT) most of the cpu time was consumed
by the method java.util.ZipFile.open(). As I traced down the thread
the first "core" method from cocoon I hit in this tree of method invocations
was the process() method of the CachingPipline. Ok that's plausibly of
course.
But the load is caused through the ClassLoaders, which I suspect unzip the
jar files of the libaries, which are used from the different serializer,
transformers,... Is this inevitabile? Or can some caching of already loaded
classes help on this problem, if it is possible? Or is this a general java
problem?

Attached are the report file of my load test and the tracing of
java.util.ZipFile.open().
I don't want to blame anybody, I just want to participate somehow.

Nice evening

Cheers
Gerhard
>Cocooners,
>
>Yesterday I fixed to major memory leaks:
> - one is with SitemapSource not releasing pipelines
>   (no memory leaks in aggregation!),
> - another with mrulist in MRUMemoryStore.
>
>Right now I have CLI running for >12 hours, it processed >35000 pages,
>and still no trace of memory problems.
>
>When doing your tests (esp. load tests), do not forget to adjust
>pool sizes (see Berin's response on "LoadTest" thread), and also
>please make sure that "heapsize" attribute of MRUMemoryStores is
>between -Xms and -Xmx.
>
>Vadim
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: cocoon-dev-unsubscribe@xml.apache.org
>For additional commands, email: cocoon-dev-help@xml.apache.org
>

Mime
View raw message