Worth trying with "none.opt" and "fixed.opt".
My guess is that "none.opt" will show a difference in the speed of the
rdf:type quesry and not for the other one.
Andy
(the overall query counts would be useful as well).
On 08/11/17 10:37, Mikael Pesonen wrote:
Hi,
only opt file is the stats.opt, and I made sure there was not such file
before running the tool.
Br
On 8.11.2017 11:13, Andy Seaborne wrote:
Mikael,
Did the database directory have a stats.opt file in it to start with?
Reading back through this thread, I can't see mention of whether there
was one or not.
An experiment on the different optimization approaches:
Take a copy of the database directory when Fuseki is not running.
In the copied TDB directory, look for all the *.opt files, note
whether there are multiple such files. Move them all elsewhere.
Put in the stats you showed "stats.opt".
Now try your queries using tdbquery --loc=... --file ...
Delete the stats file, and put in am empty file called "none.opt" and
try the queries again.
Same again for "fixed.opt": delete the "opt" file and put in an empty
one "fixed.opt".
The three files for the reordering step are "stats.opt" (with
statistics), "fixed.opt" (a built-in reordering) and ""none.opt"
(don't change the order of triples). There should only be one or zero
in the database directory; the default is "fixed".
Andy
