Base Simulation.....BufferPool Level

Results of Simulation for Buffer Pool..........BP0

Bpool GetP total........................2,379,499

Bpool Size GetP used Num. of Hits ApHit Ratio

10,000 2,362,768 1,971,692 83.5 %

18,000 2,349,204 2,049,084 87.3 %

26,000 2,322,343 2,067,504 89.1 %

34,000 2,294,862 2,070,271 90.3 %

42,000 2,267,584 2,064,701 91.1 %

50,000 2,233,613 2,044,257 91.6 %

58,000 2,171,257 1,995,045 91.9 %

66,000 2,160,792 1,991,388 92.2 %

Bpool Size Pages Read Read I/O SyHit Ratio

10,000 438.5 /S 122.6 /S 34.0 %

18,000 382.8 /S 95.0 /S 42.6 %

26,000 358.2 /S 83.1 /S 46.7 %

34,000 338.2 /S 75.1 /S 49.9 %

42,000 326.4 /S 69.2 /S 51.6 %

50,000 322.3 /S 66.3 /S 52.5 %

58,000 322.6 /S 64.2 /S 52.6 %

66,000 315.8 /S 62.1 /S 53.6 %

Previous slide Next slide Back to the first slide View Graphic Version

Notes:

Now we get to the difficult part...attempting to simulate, or estimate, pool and object performance as sizes are changed and objects are moved around.

Simulations are extremely difficult to write, and require a lot of formula ‘tweeking’ to get correct results.

However, it’s the only way to ever get things done correctly (optimally) since you can’t play dangerous and time consuming games with your production environment.

Here we see simulation results indicating increasing pool hit ratios as the size is increased. From these figures, 7,000 buffers is probably optimal unless you have memory to spare and wish to pick up that last .8%.