Okay maybe you guys didn't follow up my previous mails. I give a brief intro again.
I'm using SQLite in my program, and I have 2 threads. One thread performs insert, and the other thread perform update. The number below is the table showing that how many rows are inserted in given time. I expect that my inserts should be linear, that is, the first 30 seconds 7500, then the next should be 15000, 22500, 30000, ... However, I don't see the trend as I expected in 2.8.6. That makes me ask my question here. I've done an ADO.Net version before, so I list the results of ADO.Net is listed here as reference. If you guys has ideas about how could I tune the SQLite to make it linear, thank you I really appreciated. If you wanna tell me how to do benchmark, sorry guys, I'm not interested because that doesn't solve my problem. -Arthur -----Original Message----- From: Bert Verhees [mailto:[EMAIL PROTECTED] Sent: Saturday, November 15, 2003 2:52 AM To: [EMAIL PROTECTED] Subject: Re: [sqlite] Performance tuning question On 14 Nov 2003 at 16:28, Arthur Hsu wrote: > Hello, > > I managed to download new CVS versions (by hand ...) and compile them > on > Win32 platform (finally). Following are my test results: > > Time elapsed ADO.Net SQLite 2.8.6 SQLite CVS > 0 0 0 0 > 30 7419 5729 7920 > 60 14176 8013 10711 > 90 20760 9869 13147 > 120 26623 11033 14944 > 150 32862 12633 16598 > 180 38783 13044 17878 > 210 44472 13098 19609 > 240 49873 14120 20711 > > The CVS version is quite linear after the first 30 seconds. I'm still > digging the reason for the performance difference between 0-30 and 30-60. > Any ideas? > An important difference is that ADO.net writes records to a memory-cache, and at the point of updatedatebase writes the records to the database in a batch. If this is an advantage or not, that is a point of discussion. In a single-user environment this is mostly not a problem, except when the computer crashes, or electricity fails. But the problem with this specific benchmark is that you are comparing things you should not compare. Maybe there is a possibility to write SQLite-data to a memory cache to, and after finishing collecting data, write them to the database. Kind regards Bert Verhees > -Arthur > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]