On 28/10/11 14:31, Bruce Craig wrote:
Hello
Not a java person much yet either so pardon the naivity.
I found the tdb_init file in TDBROOT/bin
I set the JVM_ARGS to be -Xmx1200m -Xms512m
I did it like this
## ---- JVM Args e.g. -Xmx1200M
JVM_ARGS=${JVM_ARGS:-Xms512m -Xmx1200m -server}
Looks right.
I find the following in the fuseki-server script
exec java -Xmx1200M -jar "$JAR" "$@"
So it should be using the heap of 1200M there at least and that runs out of
memory as well.
I should say this generated dataset is more than 500MB on disk. Hence
asking about smaller benchmark datasets.
Thanks a bunch for any help
Bruce
Another isseu you may run into is that some VMs have simply appalling
I/O and swap performance.
But I'm still unclear as to whether the java is 32bit or 64 bit. What does
java -version
say?
TDB does different things on 32 vs 64 bit because on 64 bit you can
address the whole of large memory mapped files, and so the (highly
tuned) OS takes care of paging/caching.
Andy
