50M+ queries

Hi, In no way should this be considered definitive, but I created a table of 100 million 240-byte rows, then ran a single ESQL/C "loop 50M times and select a row into program variables on the key at random". That took 13 minutes 26 seconds. I then ran 5 copies of the program each looping 10M times. The "slowest" one took 6 minutes 49 seconds. I then ran 97 copies of the program each looping 0.5M times. The slowest one took 7 minutes 26 seconds. (I wanted to run 100, but I couldn't get them all to kick off and I wasn't going to bugger around with the kernel for this... :-) When jobs started completing, I kicked off the three that hadn't started in the loop, they all completed within 50 seconds. Please bear in mind that this was on a relatively underconfigured two-way box running about half a dozen instances of different databases. (SG, I managed to lose your e-ddress after all that, sorry. Hope you get this anyway!) ________________________________________________________________________ Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com

None of the following numbers are official. Just random numbers inserted between randomly choosen words. Were you running with an shm connection? At 5 copies, if you were going acrossed the net, you would be exceeding 30MB/sec which even a 100 base-T couldn't handle. 50MB * 240bytes / 409secs I've run similar tests with smaller row sizes(36) and smaller tables (65536 rows). On a 10 processor(9 CPU VPs) 250MHZ Sun box I've sustained 12000 selects per second, 43 million per hour. On a 24 processor(16 CPU VPs) 250 MHZ Sun box I've done 16000 selects per second, ~58 million per hour. This was over a 100 base-T network. Given this figures I suspect a newer Sun box with 10 to 16 CPUs at 400 MHZ should meet his needs. If there are a lot of users you should certainly consider MaxConnect. And a fast network. Obnoxio The Clown wrote: > Hi, > > In no way should this be considered definitive, but I created a table of 100 > million 240-byte rows, then ran a single ESQL/C "loop 50M times and select a > row into program variables on the key at random". That took 13 minutes 26 > seconds. > > I then ran 5 copies of the program each looping 10M times. The "slowest" one > took 6 minutes 49 seconds. > > I then ran 97 copies of the program each looping 0.5M times. The slowest one > took 7 minutes 26 seconds. (I wanted to run 100, but I couldn't get them all > to kick off and I wasn't going to bugger around with the kernel for this... > :-) When jobs started completing, I kicked off the three that hadn't started > in the loop, they all completed within 50 seconds. > > Please bear in mind that this was on a relatively underconfigured two-way > box running about half a dozen instances of different databases. > > (SG, I managed to lose your e-ddress after all that, sorry. Hope you get > this anyway!) > ________________________________________________________________________ > Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com