Ok, but you might have to run this a bit longer (or more iterations). My first 10 runs give the following results (in seconds):
1 15,681 2 16,010 3 16,093 4 16,168 5 16,057 6 16,585 7 16,114 8 16,596 9 16,115 10 16,270 Jumping around a bit but it seems linear (but it's not). I'm not saying that 3.6.13 behaves the stame as 3.6.14 but this only becomes really apparent when run over an extended period of time. 3.6.15 looks promising though in this respect but looking at the first 10 runs of that version does definitely not say the whole story (at least in my case): 1 17,874 2 21,753 3 23,5 4 26,591 5 27,925 6 29,682 7 29,807 8 31,944 9 32,422 10 34,144 Increasing like crazy (but seems to level off later at 40 seconds). JP -----Original Message----- From: sqlite-users-boun...@sqlite.org [mailto:sqlite-users-boun...@sqlite.org] On Behalf Of P Kishor Sent: 16. júní 2009 16:55 To: General Discussion of SQLite Database Subject: Re: [sqlite] Database inserts gradually slowing down On Tue, Jun 16, 2009 at 10:57 AM, Jens Páll Hafsteinsson<j...@lsretail.com> wrote: > The key factor here is not the total time taken to perform these operations > but the fact that the time is *increasing* for each run. I am looking for > consistency in that I need to be able to let the application perform these > steps in constant time over a long period of time (months). > Using SQLite 3.6.13, Perl 5.8.8, on a Mac OS X Leopard Macbook with 4 GB RAM and 320 GB 7200 RPM Seagate drive, I get [11:27 AM] ~/Projects/sqlite_test$perl db_slows.pl 1. the code took: 8 wallclock secs ( 6.83 usr + 0.30 sys = 7.13 CPU) 2. the code took: 9 wallclock secs ( 6.78 usr + 0.34 sys = 7.12 CPU) 3. the code took: 8 wallclock secs ( 6.80 usr + 0.34 sys = 7.14 CPU) 4. the code took: 8 wallclock secs ( 6.78 usr + 0.34 sys = 7.12 CPU) 5. the code took: 8 wallclock secs ( 6.78 usr + 0.33 sys = 7.11 CPU) 6. the code took: 8 wallclock secs ( 6.81 usr + 0.33 sys = 7.14 CPU) 7. the code took: 8 wallclock secs ( 6.80 usr + 0.34 sys = 7.14 CPU) 8. the code took: 9 wallclock secs ( 6.80 usr + 0.35 sys = 7.15 CPU) 9. the code took: 8 wallclock secs ( 6.79 usr + 0.34 sys = 7.13 CPU) 10. the code took: 8 wallclock secs ( 6.90 usr + 0.34 sys = 7.24 CPU) Pretty linear performance. As stated in the OP, I 1. Insert 1000 records (an integer and a 100 random char string; 2. Commit; 3. Repeat 1 one hundred times; 4. DELETE all records; 5. Repeat 1-4 x 10. The table is CREATE TABLE foo (id INTEGER PRIMARY KEY, num INTEGER, desc TEXT); There are no indexes. man/listinfo/sqlite-users _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
