2011/11/9 GB <gbi...@web.de> > Maybe you could try to use a pagesize that matches the size of a disk > allocation unit or memory page. For Windows since a typical NTFS partition > has a clustersize of 4KiB - which happens to also be the size of a memory > page - a pagesize of 4096 Byte seems to be a good compromise between > pleasing the cache manager and waste of space.
I already used a page_size of 4096 for all the benchmarks, together with a large (300MB) cache_size, but thanks for the suggestion anyway! > You could also investigate the effect of VACUUMing and ANALYZEing the > Database. If you have multiple indexes on tables try compiling sqlite > defining the SQLITE_ENABLE_STAT3 switch to get even better results from > ANALYZE. > I'm running these tests on a very simple database: 1 table and 1 column, so ANALYZE shouldn't have any effect. And I already tested running VACUUM on the database, but it didn't help (which seems logical, because I start with a fresh db each time). _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
