Created a table with two columns, id (number) and spellednumber (varchar). Inserted 1 million rows, each row with id containing a serial number starting at 1, and spellednumber containing the same number spelled out in English.
Then I selected 100 numbers at random from the table, and ran the test 1000 times (so, 100k SELECTs).
SQLServer was being accessed over the LAN. While it had pretty favorable cpu seconds and rate (rate is transactions/sec), it really got nailed on wallclock (elapsed) seconds... it would just max the cpu out... I think it was probably because of the log files it feels compelled to create. As such, I did only 100x100 SELECTs with SQLServer else it would have been cranking for half a day. Also, its transaction rate dropped as the number of SELECTs went up.
Rather simplistic, but telling. SQLite shines brightly, so, Richard, feel free to add it to the web site if you so desire. Thanks for a great little program.
Here are the results with the best on top. INSERT 1,000,000 db wallclock usr sys cpu rate filesize SQLite 115.00 110.02 1.75 111.77 89.47 76,736 Bdb 266 90.92 32.67 123.59 80.91 85,584 SQLServer 3119.00 206.83 79.63 286.46 34.91 74,861 Access 826.00 773.66 46.09 819.75 12.20 185,660
SELECT 100,000 db wallclock usr sys cpu rate Bdb 119.00 3.36 2.34 5.70 175.32 SQLite 23.00 13.38 9.48 22.86 43.74 SQLServer 2674.00 2.72 1.36 4.08 24.52 100tests Access 95.00 37.36 20.27 57.63 17.35
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
