As promised the follow up with some test results. I created CLEREZZA-736 [1] and attached the test report and a patch
Hopefully others have some ideas why the new implementation still consumes so much memory. Regards, Minto [1] https://issues.apache.org/jira/browse/ Op 8-3-2013 12:09, Minto van der Sluis schreef: > Hi Reto, > > I already modified the existing implementation. Later I will post my > changes and results. Now I am kind of in a hurry. > > Regards > > Minto > > Op 8-3-2013 8:51, Reto Bachmann-Gmür schreef: >> Not sure why this would relevantly increase performance. Access to set is >> supposed to be fast. >> > This is not for performance reasons but for scalability. Especially for > very large indexes it should not be required to keep the full index in > memory all the time. In my implementation I chose to use a specific > named graph for the index. > > It appear that it does increase performance in the long run, but that > will be in my full report later. > >> Good, then let's optimize the clerezza perfomance. I think the Sparql >> fastlane is an important step. Reto > Like mentioned I already started. Also did some performance comparison > between the single dataset provider and my modified version. > > Full report comes later. > > > -- ir. ing. Minto van der Sluis Software innovator / renovator Xup BV Mobiel: +31 (0) 626 014541
