OK, I will try it with 100M. Nevertheless, the diffs get built by the
recommender startup automatically. (standard behaviour - I took the
grouplens example recommender and just modified the data model)
Am 26.02.2011 12:44, schrieb Sean Owen:
I don't know enough about your code to guess at an explanation. I don't know
whether you were actually building the diffs before as it sounds like you
were looking to build it manually or something.
I am not sure if 3GB is enough, or if it is, whether you want to spend that
much memory on diffs. You could start with a lower number like 10M.
2011/2/26 Daniel Mühlbachler<[email protected]>
Ok, that's true because I in my preferenceTable I have about 650 000
entries/ratings (~24MB) but it's curious that it worked before perfectly
without any problem and I have nothing changed on my system except the svn
update and recompilation. It also takes approximately as long as before to
build the diffs.
So, you think if I say 100 million entries should be kept about 3GB of heap
size is enough?
