The PROMON gather script shows that the database is checkpointing every 2 seconds.
Checkpoints:
Ckpt ------ Database Writes ------
No. Time Len Freq Dirty CPT Q Scan APW Q Flushes Duration Sync Time
20484 12:43:43 0 0 6 5 0 0 0 0.42 0.00
20483 12:43:41 0 2 29 29 0 0 0 0.81 0.00
20482 12:43:37 4 4 6 3 0 0 2 0.60 0.00
20481 12:43:33 0 4 5 4 0 0 1 0.41 0.00
20480 12:43:24 4 9 6 4 0 0 0 0.62 0.00
20479 12:43:22 2 2 35 31 0 0 2 0.67 0.00
20478 12:43:21 1 1 38 35 0 0 2 0.74 0.00
20477 12:43:19 0 2 4 3 0 0 1 0.43 0.00
The
bi cluster size is too small which results in too frequent checkpointing. When checkpointing this frequently, the APWs cannot write out all the modified buffers. This results in having hard flushes. When this occurs, all processes wait until the modified buffers have been written to disk.
