https://issues.apache.org/bugzilla/show_bug.cgi?id=53508
--- Comment #2 from [email protected] --- Hello Yegor Thank you for very fast response! I must admit, that I tried to create a testcase, but failed. But problem on our server persisted. I recognized that the problem disappeared only with server restart. First I recognized this on our testing server. Then I proofed this a second time at our production server. I was able to measure the difference: Long running server performance was around 30 times slower than fresh started server. After some over night XMLX-Exports on our production server the performance is down again (perhaps 20 times slower). Test server has had no workload and is unchanged fast. We figured out only with thread dumps, that the SharedStringsTable Hash Key Generation is perhaps the problem, because threads was hanging in this method very often. Obviously the POI code causes not the 30 times slowdown in performance. Perhaps there is a bad cache implementation or something else in the XML-Beans code? Currently I have not tested your code snippet. Surely this could help to speed up POI anyway. I will try to search something further.. Before Server restart: 2012-07-04 16:44:42,423 ERROR table.Table.serializeXLS() (729) - serializeXLS for 31681 rows with 20 columns for user 'XXX' lasted 473392 with an avg of 14.942457624443673 per row After Server restart: 2012-07-04 18:56:29,686 ERROR table.Table.serializeXLS() (729) - serializeXLS for 31681 rows with 20 columns for user 'XXX' lasted 14919 with an avg of 0.4709131656197721 per row After some over night export workload this morning: 2012-07-05 10:52:00,446 ERROR table.Table.serializeXLS() (729) - serializeXLS for 31680 rows with 20 columns for user 'XXX' lasted 281254 with an avg of 8.877967171717172 per row (Each of this durations was stable reproducible over several test export runs) Best wishes Daniel -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
