system slowing down after several runs

TheBearTheBear Member Posts: 18 Maven
edited November 2018 in Help
I experience (especially with 4.3) a system slow down when I am running
several times with (nearly) the same operator structur on the same data set.
Sometimes it says HeapBuffer full or alike.. But most often it just takes ages to start the process.
When I restart RM everything runs smoothly.

My dataset contains only 15.000 examples and 20 attributes. I have 1 GB of RAM and 1.8GHz.
Operators are just some simple preprocessors like numerical2polinomial as well as IOmultiplier (just 1 or two copies)and NaiveBayes. But most often I terminate the runs in the preprocessing strucure...

I use a lot the breakpoints and change the order of operators slightlywith heaps of syntax errors and warnings...I just use trial and error to get deeper insights in the paradigm of RM. Maybe this is a reason for the slowing down? Anyone else experience these behaviour?

(I also tried MemoryCleanUp but it does not show any effect)

Answers

  • IngoRMIngoRM Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Community Manager, RMResearcher, Member, University Professor Posts: 1,751 RM Founder
    Hi,

    There seems to be a GUI related memory leak in cases where results are displayed at a breakpoint or at the end of the process. In many cases, the resources are not freed (at least not for a long time) after the results are displayed once. The whole team is currently profiling RapidMiner and searching for this leak but as I have to admit we did not have any success yet. I will let you know as soon as we found the reason and will deliver an updated version as soon as possible.

    The effect is of course stronger if you use a lot of breakpoints and hence a lot of visualizations. By the way: the slowdown is probably a result of swapping when the memory starts to fill.

    We will keep you updated and will release an update as soon as possible.

    Cheers and thanks for pointing this out,
    Ingo

Sign In or Register to comment.