Analyzing effects of different buffer pool attributes

The following steps show the effects of using object placement and simulation, based on the information found in the activity reports.

  1. Performing a simulation using the actual object placements:
    1. The four most active buffer pools BP1, BP2, BP3, and BP10 are chosen.
    2. A minimum and maximum page size for all buffer pools of 25 000 and 1 000 000 and a simulation interval of 25 000 is chosen (the actual buffer pool size for BP2 and BP3 is approximately 300 000 pages each).
    3. The object to buffer pool assignments remain unchanged (the actual object placements are to be simulated).
    Table 1. Comparing separate Buffer Pools versus a single combined Buffer Pool
      Separate Buffer Pools Combined Buffer Pool
    Total Pages Misses Application Hit Ratio Global Miss Ratio Misses Application Hit Ratio Global Miss Ratio
    100000 2418999 79.1 20.9 463512 96.0 4.0
    125000 880442 92.4 7.6 436713 96.2 3.8
    150000 464793 96.0 4.0 412713 96.4 3.6
    175000 439478 96.2 3.8 389354 96.6 3.4
    200000 415460 96.4 3.6 366904 96.8 3.2
    225000 397097 96.6 3.4 353614 96.9 3.0
    250000 376870 96.7 3.3 333667 97.1 2.9
    275000 354308 96.9 3.1 319326 97.2 2.8
    300000 333677 97.1 2.9 308144 97.3 2.7
    325000 317252 97.3 2.7 298350 97.4 2.6
    350000 305352 97.4 2.6 290182 97.5 2.5
    375000 295169 97.5 2.5 284812 97.5 2.5
    400000 287248 97.5 2.5 277367 97.6 2.4
    425000 280833 97.6 2.4 273104 97.6 2.4
    450000 274582 97.6 2.4 268341 97.7 2.3
    500000 264242 97.7 2.3 257447 97.8 2.2
    ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮
    975000 210329 98.2 1.8 208484 98.2 1.8
    1000000 209556 98.2 1.8 208140 98.2 1.8
    • For a total buffer pool size of less than 300 000 pages the hit ratio is better if BP2 and BP3 are combined in a single buffer pool. Above 300 000 pages the improvement is marginal. An object placement should be performed to determine if there is a more favorable distribution of objects.
    • It is noticeable that even around 1 000 000 pages the buffer pool hit ratio still improves slowly with further memory.
    Table 2. Recommended sizing for separate Buffer Pools
    Total Pages BP1 pages BP2 pages BP3 pages BP10 pages
    100000 25000 25000 25000 25000
    125000 25000 50000 25000 25000
    150000 25000 50000 50000 25000
    175000 25000 75000 50000 25000
    200000 25000 100000 50000 25000
    225000 25000 125000 50000 25000
    250000 25000 150000 50000 25000
    275000 25000 175000 50000 25000
    300000 25000 200000 50000 25000
    325000 25000 225000 50000 25000
    350000 25000 250000 50000 25000
    375000 25000 275000 50000 25000
    400000 25000 275000 75000 25000
    425000 25000 300000 75000 25000
    450000 25000 325000 75000 25000
    475000 25000 325000 100000 25000
    500000 25000 350000 100000 25000
    525000 25000 375000 100000 25000
    550000 25000 425000 75000 25000
    575000 25000 425000 100000 25000
    600000 25000 425000 125000 25000
    625000 25000 425000 150000 25000
    650000 25000 425000 175000 25000
    675000 25000 450000 175000 25000
    700000 25000 450000 200000 25000
    725000 25000 450000 225000 25000
    750000 25000 600000 100000 25000
    775000 25000 575000 150000 25000
    800000 25000 600000 150000 25000
    825000 25000 600000 175000 25000
    850000 25000 625000 175000 25000
    875000 25000 650000 175000 25000
    900000 25000 650000 200000 25000
    925000 25000 675000 200000 25000
    950000 25000 675000 225000 25000
    975000 25000 675000 250000 25000
    1000000 25000 725000 225000 25000
    • The most important result is the allocation of memory between the two main buffer pools BP2 and BP3 (each one actually using approximately 300 000 pages): For a total of 600 000 pages, the simulation recommends to allocate three times as much memory to BP2 as BP3.
    • The values for buffer pools BP1 and BP10 are wasteful and caused by the high interval value simulated (25 000 pages). An additional simulation between 500 and 5 000 pages, only for BP0, BP1, BP4 and BP10, will give a better idea of the amount of memory they really require, but this is absolutely insignificant compared to BP2 and BP3.
    Simulated behavior of each separate Buffer Pool of the simulation result shows:
    Table 3. Results of simulated behavior of separate buffer pools
      Buffer Pool BP1 Buffer Pool BP2 Buffer Pool BP3 Buffer Pool BP10
    Buffer Pool Pages Misses Applic. Hit Ratio Global Miss Ratio Misses Applic. Hit Ratio Global Miss Ratio Misses Applic. Hit Ratio Global Miss Ratio Misses Applic. Hit Ratio Global Miss Ratio
    25000 359 99.8 0.0 1909336 67.3 16.5 509028 90.8 4.4 276 98.1 0.0
    50000 359 99.8 0.0 370779 93.6 3.2 93379 98.3 0.8 276 98.1 0.0
    75000 359 99.8 0.0 345464 94.1 3.0 85458 98.5 0.7 276 98.1 0.0
    100000 359 99.8 0.0 321446 94.5 2.8 79923 98.6 0.7 276 98.1 0.0
    125000 359 99.8 0.0 303083 94.8 2.6 76671 98.6 0.7 276 98.1 0.0
    150000 359 99.8 0.0 282856 95.2 2.4 7352 98.7 0.6 276 98.1 0.0
    175000 359 99.8 0.0 260294 95.5 2.2 70005 98.7 0.6 276 98.1 0.0
    200000 359 99.8 0.0 239663 95.9 2.1 68202 98.8 0.6 276 98.1 0.0
    225000 359 99.8 0.0 223238 96.2 1.9 67029 98.8 0.0 276 98.1 0.0
    250000 359 99.8 0.0 211338 96.4 1.8 66103 98.8 0.6 276 98.1 0.0
    275000 359 99.8 0.0 201155 96.6 1.7 65542 98.8 0.6 276 98.1 0.0
    300000 359 99.8 0.0 194740 96.7 1.7 63988 98.8 0.6 276 98.1 0.0
    325000 359 99.8 0.0 188489 96.8 1.6 63473 98.9 0.5 276 98.1 0.0
    350000 359 99.8 0.0 183684 96.9 1.6 63102 98.9 0.5 276 98.1 0.0
    375000 359 99.8 0.0 179376 96.9 1.5 62749 98.9 0.5 276 98.1 0.0
    400000 359 99.8 0.0 174842 97.0 1.5 62528 98.9 0.5 276 98.1 0.0
    425000 359 99.8 0.0 168940 97.1 1.5 62024 98.9 0.5 276 98.1 0.0
    450000 359 99.8 0.0 165805 97.2 1.4 60979 98.9 0.5 276 98.1 0.0
    475000 359 99.8 0.0 164905 97.2 1.4 60853 98.9 0.5 276 98.1 0.0
    500000 359 99.8 0.0 163878 97.2 1.4 60850 98.9 0.5 276 98.1 0.0
    525000 359 99.8 0.0 162486 97.2 1.4 60850 98.9 0.5 276 98.1 0.0
    550000 359 99.8 0.0 159330 97.3 1.4 60850 98.9 0.5 276 98.1 0.0
    575000 359 99.8 0.0 154300 97.4 1.3 60850 98.9 0.5 276 98.1 0.0
    600000 359 99.8 0.0 150934 97.4 1.3 60850 98.9 0.5 276 98.1 0.0
    625000 359 99.8 0.0 148159 97.5 1.3 60850 98.9 0.5 276 98.1 0.0
    650000 359 99.8 0.0 145146 97.5 1.3 60850 98.9 0.5 276 98.1 0.0
    675000 359 99.8 0.0 143591 97.5 1.2 60850 98.9 0.5 276 98.1 0.0
    700000 359 99.8 0.0 143263 97.5 1.2 60850 98.9 0.5 276 98.1 0.0
    725000 359 99.8 0.0 141892 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    750000 359 99.8 0.0 141783 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    775000 359 99.8 0.0 141569 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    800000 359 99.8 0.0 141491 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    825000 359 99.8 0.0 141377 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    850000 359 99.8 0.0 141331 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    875000 359 99.8 0.0 141281 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    900000 359 99.8 0.0 141275 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    925000 359 99.8 0.0 141275 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    950000 359 99.8 0.0 141275 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    975000 359 99.8 0.0 141275 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    1000000 359 99.8 0.0 141275 97.6 1.2 60850 98.9 0.5 276 98.1 0.0
    • The Misses columns for buffer pools BP2 and BP3 show that, with unlimited memory, no improvement would be achieved with more than 900 000 pages for BP2 and 500 000 pages for BP3. These are the absolute limits, although economical limits will always be lower.

    The other tables in this simulation result are only required for very detailed analysis, especially for application tuning.

  2. Performing object placement with default rule set and object placements:
    1. The default rule set is used; in this case pattern_large.
    2. No changes are made to rules or placements. The object placement wizard is started by simply clicking Next.
    ALTER BUFFERPOOL COMMANDS of the object placement result shows the new buffer pools and the recommended sizes: 
    ALTER BUFFERPOOL(BP3) VPSIZE(119555) HPSIZE(0) VPSEQT(20) DWQT(39) VDWQT(10,0)
ALTER BUFFERPOOL(BP2) VPSIZE(392018) HPSIZE(0) VPSEQT(20) DWQT(0) VDWQT(0,0)
ALTER BUFFERPOOL(BP10) VPSIZE(15046) HPSIZE(0) VPSEQT(40) DWQT(10) VDWQT(3,0)
ALTER BUFFERPOOL(BP0) VPSIZE(15046) HPSIZE(0) VPSEQT(20) DWQT(25) VDWQT(6,0)
ALTER BUFFERPOOL(BP32K) VPSIZE(1881) HPSIZE(0) VPSEQT(40) DWQT(10) VDWQT(3,0)
ALTER BUFFERPOOL(BP4) VPSIZE(15046) HPSIZE(0) VPSEQT(98) DWQT(2) VDWQT(0,0)
ALTER BUFFERPOOL(BP1) VPSIZE(15046) HPSIZE(0) VPSEQT(100) DWQT(70) VDWQT(50,0)
    ALTER TABLESPACE AND ALTER INDEX STATEMENTS of the object placement result shows the object placement commands (for those objects being allocated to a different buffer pool): 
    ALTER TABLESPACE WTNTEST.WTNADD BUFFERPOOL BP3;
ALTER INDEX WTNTEST.WTNADR01 BUFFERPOOL BP2;
ALTER TABLESPACE WTNTEST.WTNARC BUFFERPOOL BP3;
ALTER TABLESPACE WTNTEST.WTNHST BUFFERPOOL BP3;
ALTER INDEX WTNTEST.WTNTRD01 BUFFERPOOL BP2;
ALTER INDEX WTNTEST.WTNAST01 BUFFERPOOL BP2;
ALTER INDEX WTNTEST.WTNBAL01 BUFFERPOOL BP2;
    • The actual object assignments (at data collection time) separate data from indexes.
    • The default rule set in the selected pattern file has separated objects primarily according to the amount of sequential and dynamic access.
    OBJECT PLACEMENT OVERVIEW of the object placement result shows attributes and placement of objects:
    Table 4. Results of object placement
    Object Name Type Page Used Catalog / Directory Seq. Access [%] Change Rate [%] Size [pages] Current Recommended User- defined
    WTNTEST.DBD01 TABLESPACE 4K YES DIR 0 133 ? BP0 BP0 BP0
    WTNTEST.DSNLLX01 INDEX 4K YES DIR 0 6 ? BP0 BP0 BP0
    WTNTEST.DSNLLX02 INDEX 4K YES DIR 0 28 ? BP0 BP0 BP0
    WTNTEST.DSNLUX01 INDEX 4K YES DIR 0 9 ? BP0 BP0 BP0
    WTNTEST.DSNLUX02 INDEX 4K YES DIR 0 21 ? BP0 BP0 BP0
    ⋮                    
  3. Performing a simulation with results from object placement:
    1. The two most active buffer pools BP2 and BP3 are chosen (they contain almost all database activity).
    2. A minimum and maximum page size for all buffer pools of 100 000 and 1 000 000 and a simulation interval of 25 000 is chosen.
    Simulated behavior of each separate Buffer Pool of the simulation result shows:
    Table 5. Results of simulated behavior of most active buffer pools
      Buffer Pool BP2 Buffer Pool BP3
    Buffer Pool Pages Misses Application Hit Ratio Global Miss Ratio Misses Application Hit Ratio Global Miss Ratio
    100000 437337 96.0 3.8 12904 97.9 0.1
    125000 408394 96.3 3.5 12886 97.9 0.1
    150000 366438 96.5 3.3 12886 97.9 0.1
    175000 362053 96.7 3.1 12886 97.9 0.1
    200000 341920 96.9 2.9 12886 97.9 0.1
    225000 325018 97.0 2.8 12886 97.9 0.1
    250000 305094 97.2 2.6 12886 97.9 0.1
    275000 293311 97.3 2.5 12886 97.9 0.1
    300000 282036 97.4 2.4 12886 97.9 0.1
    325000 273809 97.5 2.4 12886 97.9 0.1
    350000 266564 97.6 2.3 12886 97.9 0.1
    375000 259972 97.6 2.2 12886 97.9 0.1
    400000 250910 97.7 2.2 12886 97.9 0.1
    ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮
    • The application hit ratio for buffer pool BP2 is marginally better than in the previous simulation (look at misses for a given number of total buffer pool pages).
    • Buffer pool BP3 is already optimal at 100 000.
  4. Performing object placement with modified rule set:
    1. The default rule set pattern_large is selected and edited as follows:
      1. A duplicate BP2, called BP12, is added directly under BP2.
      2. A duplicate BP3, called BP13, is added directly under BP3.
      3. BP2 and BP3 should only be used for table spaces.
      4. BP12 and BP13 should only be used for index spaces.
    2. The edited pattern file is saved with a new name. This pattern will be chosen as the default for all future placements with this subsystem.
    3. No changes are made to placements.
  5. Performing simulation with results from object placement with modified rule set:
    1. Buffer pools BP2, BP3, BP12, and BP13 are chosen.
    2. A minimum and maximum page size for all buffer pools of 25 000 and 1 000 000 and a simulation interval of 25 000 is chosen.
    Recommended sizing for separate Buffer Pools of the simulation result shows:
    Table 6. Recommended sizing for separate buffer pools with modified rule set
    Total Pages BP2 pages BP3 pages BP12 pages BP13 pages
    100000 25000 25000 25000 25000
    125000 50000 25000 25000 25000
    150000 50000 25000 50000 25000
    175000 75000 25000 50000 25000
    200000 100000 25000 50000 25000
    225000 125000 25000 50000 25000
    250000 150000 25000 50000 25000
    275000 175000 25000 50000 25000
    300000 200000 25000 50000 25000
    325000 225000 25000 50000 25000
    350000 250000 25000 50000 25000
    375000 275000 25000 50000 25000
    400000 300000 25000 50000 25000
    425000 300000 25000 75000 25000
    450000 300000 25000 100000 25000
    475000 325000 25000 100000 25000
    500000 350000 25000 100000 25000
    525000 350000 25000 125000 25000
    550000 400000 25000 100000 25000
    575000 425000 25000 100000 25000
    600000 425000 25000 125000 25000
    625000 425000 25000 150000 25000
    650000 425000 25000 175000 25000
    675000 425000 25000 200000 25000
    700000 550000 25000 100000 25000
    725000 575000 25000 100000 25000
    750000 575000 25000 125000 25000
    775000 600000 25000 125000 25000
    800000 625000 25000 125000 25000
    825000 625000 25000 150000 25000
    850000 625000 25000 175000 25000
    875000 650000 25000 175000 25000
    900000 650000 25000 200000 25000
    925000 675000 25000 200000 25000
    950000 675000 25000 225000 25000
    975000 675000 25000 250000 25000
    1000000 650000 25000 300000 25000
    • Most of the memory is allocated to BP2 followed by BP12.
    Simulated behavior of each separate Buffer Pool of the simulation result shows:
    Table 7. Results of simulated behavior of separate buffer pools with modified rule set
      Buffer Pool BP2 Buffer Pool BP3 Buffer Pool BP12 Buffer Pool BP13
    Buffer Pool Pages Misses Applic. Hit Ratio Global Miss Ratio Misses Applic. Hit Ratio Global Miss Ratio Misses Applic. Hit Ratio Global Miss Ratio Misses Applic. Hit Ratio Global Miss Ratio
    25000 1870289 66.5 16.1 3529 99.2 0.0 490692 90.8 4.2 10037 94.7 0.1
    50000 362552 93.5 3.1 3320 99.2 0.0 79115 98.5 0.7 9817 94.8 0.1
    75000 337339 94.0 2.9 3320 99.2 0.0 72303 98.6 0.6 9686 95.0 0.1
    100000 313468 94.4 2.7 3320 99.2 0.0 65437 98.8 0.6 9566 95.0 0.1
    125000 294921 94.7 2.5 3320 99.2 0.0 62273 98.8 0.5 9566 95.0 0.1
    150000 274273 95.1 2.4 3320 99.2 0.0 59767 98.9 0.5 9666 95.0 0.1
    175000 251441 95.5 2.2 3320 99.2 0.0 58299 98.9 0.5 9566 95.0 0.1
    200000 226048 95.9 1.9 3320 99.2 0.0 56990 98.9 0.5 9566 95.0 0.1
    225000 211886 96.2 1.8 3320 99.2 0.0 56176 98.9 0.5 9566 95.0 0.1
    250000 202081 96.4 1.7 3320 99.2 0.0 55614 99.0 0.5 9566 95.0 0.1
    275000 193639 96.5 1.7 3320 99.2 0.0 55162 99.0 0.5 9566 95.0 0.1
    300000 185904 96.7 1.6 3320 99.2 0.0 53560 99.0 0.5 9566 95.0 0.1
    325000 180557 96.8 1.6 3320 99.2 0.0 52788 99.0 0.5 9566 95.0 0.1
    350000 175611 96.9 1.5 3320 99.2 0.0 51485 99.0 0.4 9566 95.0 0.1
    375000 172480 96.9 1.5 3320 99.2 0.0 51319 99.0 0.4 9566 95.0 0.1
    400000 167135 97.0 1.4 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    425000 161517 97.1 1.4 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    450000 160539 97.1 1.4 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    475000 159454 97.1 1.4 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    500000 157582 97.2 1.4 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    525000 153173 97.3 1.3 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    550000 149690 97.3 1.3 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    575000 146377 97.4 1.3 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    600000 143229 97.4 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    625000 140383 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    650000 139060 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    675000 137805 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    700000 137718 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    725000 137590 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    750000 137416 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    775000 137324 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    800000 137268 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    825000 137240 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    850000 137239 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    875000 137239 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    900000 137239 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    925000 137239 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    950000 137239 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    975000 137239 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    1000000 137239 97.5 1.2 3320 99.2 0.0 51299 99.0 0.4 9566 95.0 0.1
    • The results are further improved again (look at misses for a given number of total buffer pool pages).
  6. Performing object placement with the default rule set and moving expensive objects:
    1. The default rule set pattern_large is used again (not the saved pattern from step 4).
    2. No changes are made to rules.
    3. When the object placement is performed, the identified problem table spaces and index spaces (from the detail report) are assigned to buffer pool BP10.
  7. Performing simulation with results from the object placement:
    1. Buffer pools BP2, BP3, and BP10 are chosen.
    2. A minimum and maximum page size for all buffer pools of 25 000 and 1 000 000 and a simulation interval of 25 000 is chosen.
    Recommended sizing for separate Buffer Pools of the simulation result shows:
    Table 8. Recommended sizing for separate buffer pools with moved objects
    Total Pages BP2 pages BP3 pages BP10 pages
    75000 25000 25000 25000
    100000 50000 25000 25000
    125000 50000 25000 50000
    150000 50000 25000 75000
    175000 50000 25000 100000
    200000 50000 25000 125000
    225000 50000 25000 150000
    250000 50000 25000 175000
    275000 75000 25000 175000
    300000 75000 25000 200000
    325000 75000 50000 200000
    350000 100000 50000 200000
    375000 100000 50000 225000
    400000 100000 50000 250000
    425000 100000 50000 275000
    450000 100000 50000 300000
    475000 100000 50000 325000
    500000 125000 50000 325000
    525000 150000 50000 325000
    550000 175000 50000 325000
    575000 175000 50000 350000
    600000 200000 50000 350000
    625000 175000 50000 400000
    650000 175000 50000 425000
    675000 175000 50000 450000
    700000 175000 50000 475000
    725000 200000 50000 475000
    750000 225000 50000 475000
    775000 250000 50000 475000
    800000 275000 50000 475000
    825000 300000 50000 475000
    850000 325000 50000 475000
    875000 325000 50000 500000
    900000 350000 50000 500000
    925000 375000 50000 500000
    950000 400000 50000 500000
    975000 425000 50000 500000
    1000000 450000 50000 500000
    • The memory is mainly shared between BP2 and BP10 (BP10 gets somewhat more).
    Simulated behavior of each separate Buffer Pool of the simulation result shows:
    Table 9. Results of simulated behavior of separate buffer pools with moved objects
      Buffer Pool BP2 Buffer Pool BP3 Buffer Pool BP10
    Buffer Pool Pages Misses Application Hit Ratio Global Miss Ratio Misses Application Hit Ratio Global Miss Ratio Misses Application Hit Ratio Global Miss Ratio
    25000 1977916 77.8 17.1 22826 96.3 0.2 384203 80.9 3.3
    50000 157728 98.2 1.4 13192 97.9 0.1 277984 86.2 2.4
    75000 146042 98.4 1.3 12939 97.9 0.1 255630 87.3 2.2
    100000 137354 98.5 1.2 12904 97.9 0.1 231980 88.5 2.0
    125000 132345 98.5 1.1 12886 97.9 0.1 208635 89.6 1.8
    150000 127540 98.6 1.1 12886 97.9 0.1 182152 91.0 1.6
    175000 122628 98.6 1.1 12886 97.9 0.1 148533 92.6 1.3
    200000 120617 98.6 1.0 12886 97.9 0.1 137183 93.2 1.2
    225000 119323 98.7 1.0 12886 97.9 0.1 129862 93.6 1.1
    250000 117911 98.7 1.0 12886 97.9 0.1 123232 93.9 1.1
    275000 116349 98.7 1.0 12886 97.9 0.1 117433 94.2 1.0
    300000 115100 98.7 1.0 12886 97.9 0.1 108989 94.6 0.9
    325000 113892 98.7 1.0 12886 97.9 0.1 102452 94.9 0.9
    350000 112813 98.7 1.0 12886 97.9 0.1 100332 95.0 0.9
    375000 112392 98.7 1.0 12886 97.9 0.1 98409 95.1 0.8
    400000 111509 98.7 1.0 12886 97.9 0.1 95495 95.3 0.8
    425000 111000 98.8 1.0 12886 97.9 0.1 92747 95.4 0.8
    450000 110719 98.8 1.0 12886 97.9 0.1 90016 95.5 0.8
    475000 110378 98.8 1.0 12886 97.9 0.1 87894 95.6 0.8
    500000 110040 98.8 0.9 12886 97.9 0.1 86791 95.7 0.7
    525000 109411 98.8 0.9 12886 97.9 0.1 86778 95.7 0.7
    550000 107470 98.8 0.9 12886 97.9 0.1 86778 95.7 0.7
    575000 105709 98.8 0.9 12886 97.9 0.1 86778 95.7 0.7
    600000 104477 98.8 0.9 12886 97.9 0.1 86778 95.7 0.7
    625000 103580 98.8 0.9 12886 97.9 0.1 86778 95.7 0.7
    650000 102072 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    675000 101881 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    700000 101765 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    725000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    750000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    775000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    800000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    825000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    850000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    875000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    900000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    925000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    950000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    975000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    1000000 101760 98.9 0.9 12886 97.9 0.1 86778 95.7 0.7
    • There is an improvement at 600 000 pages, but it is worse at 1 000 000 pages.

Further object placements and simulations could be performed for a combination of the saved rule set and putting problem table spaces into a separate buffer pool.