On 27/10/11 21:15, Bruce Craig wrote:
We are new to JENA* and thought we would first try to replicate benchmarks
for the tools that we have seen mentioned.
One which aligned well with the kinds of work we are pursuing was the Social
Network Intelligence Benchmark (
http://www.w3.org/wiki/Social_Network_Intelligence_BenchMark )
We downloaded the generator and created what seems like a huge dataset.
Nonetheless we succeeded in creating a TDB data store we think with
tdbloader. However, any attempts to tdbdump and query with arq or tdbquery
or connect and query with fuseki greets us with heap memory failures.
Out of heap? What heap size are you using? TDB does not flex with heap
size (in the last release).
On a 32 bit JVM, use -Xmx1200M
On a 64 bit JVM, don't push the heap size up too high (caching of
indexes isn't in the heap; nodes are though).
Seeing the stacktrace would be useful.
Andy
We dont really need DBpedia scale as yet so we were looking for experience
and guidance on
a) Setup of basic tool set - any hidden caveats?
b) Suggestions for more modest datasets to use in environments on VM linux
systems with 2-4GB RAM.
Regards
Bruce
(caution - that wiki page documents XSD datatimes as being
"2011-02-22 09:43:13"^^xsd:dateTime
they are not. TDB (nor ARQ) will recognize that as a dataTime. No
space, use a T.
e.g.
"2011-02-22T09:43:13"^^xsd:dateTime
