Here's four more points on the curve - I'd use a "dirac delta function" for your curve fit ;-)
Shared_buffers Select Count Vacuum (KB) (s) (s) ======================================= 248 5.52 2.46 368 4.77 2.40 552 5.82 2.40 824 6.20 2.43 1232 5.60 3.59 1848 6.02 3.14 2768 5.53 4.56 5536 6.05 3.95 8304 5.80 4.37 12456 5.86 4.12 18680 5.83 4.10 28016 6.11 4.46 WRT what you found on the selection algorithm, it might also explain the L2 effects I think. I'm also still of the opinion that polluting the shared buffer cache for a seq scan does not make sense. - Luke On 3/5/07 10:21 AM, "Luke Lonergan" <[EMAIL PROTECTED]> wrote: > Tom, > > On 3/5/07 8:53 AM, "Tom Lane" <[EMAIL PROTECTED]> wrote: > >> Hm, that seems to blow the "it's an L2 cache effect" theory out of the >> water. If it were a cache effect then there should be a performance >> cliff at the point where the cache size is exceeded. I see no such >> cliff, in fact the middle part of the curve is darn near a straight >> line on a log scale ... > > Here's that cliff you were looking for: > > Size of Orders table: 7178MB > Blocksize: 8KB > > Shared_buffers Select Count Vacuum > (KB) (s) (s) > ======================================= > 248 5.52 2.46 > 368 4.77 2.40 > 552 5.82 2.40 > 824 6.20 2.43 > 1232 5.60 3.59 > 1848 6.02 3.14 > 2768 5.53 4.56 > > All of these were run three times and the *lowest* time reported. Also, the > behavior of "fast VACUUM after SELECT" begins abruptly at 1232KB of > shared_buffers. > > These are Opterons with 2MB of L2 cache shared between two cores. > > - Luke > > > > > ---------------------------(end of broadcast)--------------------------- > TIP 6: explain analyze is your friend > ---------------------------(end of broadcast)--------------------------- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to [EMAIL PROTECTED] so that your message can get through to the mailing list cleanly