Number of document decrease XWiki performance

Hi all, I have until now not solved this problem (since more than 10 month): - using this program to create automatically a document work well for few documents: XW = xwiki.XWiki; XWContext = context.context; XWikiDocument NewEntry = new XWikiDocument(request.getParameter("web"), SourceValue); NewEntry.setContent(EntryContent); XW.saveDocument(NewEntry, XWContext); When putting these instructions in a LOOP for 10000 entries to transfer from another TABLE in the same XWiki database cause a problem. The probleme is not an error but the XWiki crash! and without finishing the LOOP, the xwiki display error "out of memory!" But now, taking the same program and run it in the shell (javac, java) to transfert these 10000 entries to TABLE in the XWiki database, it's done in less 1 minute? Strange for me ... - More ...and this is the more importante thing : One time I success to transfer 25 000 documents in the XWiki database (using the same script above) the XWiki become very heavy and cannot really work with it at all. MySql tables are supposed to contain millions of records, and when using the database adminstrator and check with SQL statment, I get resultat in few seconds, that means that the problem is in XWiki. My Question : Why with 25 000 documents XWiki don't work as usually Anyone faced this problem before? Do you have made testing of XWiki for 1 millions documents for exemple? I have XP, 1.2 MHz and 500 Mo but the following experience on MySQL gives more arguments that not MySQL is causing this problem : (Web link : http://www.guilde.asso.fr/lurker/message/20050817.183426.a1b8e267.en.html) ----------------------------------------------------------------- Il y a quelques années, j'ai utilisé MySQL pour gérér 1.7 millions d'enregistrements dans une table. La table faisait environ 500Mo, plus 400Mo d'indexes pour la recherche de textes dans un des champs. Coté perfs, sur un ordinateur portable à 500Mhz et 128Mo de ram qui faisait tourner un Windows NT4 bourré d'applis en mémoire, il fallait environ 5s à 10s pour recherche les tuples contenant certains mots-clefs stokés dans le fameux champs de texte. Bref, des performances tout à fait acceptable, sur une machine qui n'était pourtant pas du tout fait pour faire serveur de base de données (et je ne parle pas que de l'OS...). -------------------------------------------------------------- Many thanks for any suggestion. Best wishes ---------------------------------- Youcef Bey