Hello! В сообщении от Monday 16 February 2009 17:42:25 danjenkins написал(а): > I fully understand that performance will depend on the coding, database > structure and indexing (& hardware) but, assuming these are taken care of, > should a 100 million record table perform loosely in the same performance > class as other popular databases?
I did test SQLite databases up to 100 GB size with 1 000 000 blobs or 100 000 000 strings in single table. Of cource, write operations must be grouped becouse memory allocation for write transaction is proportional to database size (see offsite). I did not find any explicit non-linear size effects. P.S. In real projects I'm prefer to split databases of a few gigabytes pieces and attach these when it's needed. For example, a lot of datasets can be splitted by months. On debian etch/lenny hosts with ext3 filesystem and RAM>1 Gb this options may be useful: pragma page_size=4096; PRAGMA default_cache_size=200000; PRAGMA cache_size=200000; P.P.S. SQLite very effectively work with huge databases when database size >> RAM size. And databases is compact. So, 4,5 Gb SQLite database work fine but 18 Gb PostgreSQL database with equal data (yes, PostgreSQL databases are not compact) is work very bad on linux host with 1Gb RAM. Best regards, Alexey. _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
