Hi Thomas, >- Is autocommit on? Or do you commit yourself? If yes, after how much >rows?
Where do I change autocommit? Is it more efficient if I issue a commit every 500 rows? - A local connection can be done with shared memory or with the >loopback device. > If you specify "localhost" as host, shared memory is used. If you > specify > <myhostname> than loopback. The latter will be slower, of course. I just didn't specify any hostname. >- Do you insert via parameter binding or with SQL statement literals? >- Do you use prepared statements? (MySQL might have problems here) I just issue SQL statement litterals. Indeed mySQL does not plan that feature till their version 5. >> I am looking for raw INSERT speed. Data security and integrity are >> secondary factors. > >In this case probably MySQL is the right stuff for you. However, if you >change your mind about security and integrity, something data bases >are built for ... I need to save as many intermediary results as possible. If the result does not exist in the database it is recalculated. At this point it becomes clear that no DB engine seems to be able to handle one GB of data speedily without souped up hardware. Considering the cost of RAM one wonders if it would not be more efficient to simply handle DBs in RAM while logging changes on HD - on replicated machines for the sake of safety, if need be. The DB can be updated on HD as an End of day process. A replicated HD and RAM DB makes Business resumption speed a lesser issue. It is of course longer to rebuild the RAP DB from the log, but chances of two concomitent server crashes are small. Considering the opportunity for data compression, the model may have some merits even for several several GB DBs. Does LiveCache have a speedy way to save the RAM DB to file, possibly at set intervals? Does it compress data? Best, Terry _______________________________________________ sapdb.general mailing list [EMAIL PROTECTED] http://listserv.sap.com/mailman/listinfo/sapdb.general
