Oracle hidden parallel parameters Don Burleson

Note: Oracle hidden parameters are undocumented and unsupported by Oracle.
Make sure that you thoroughly test all hidden parameters before placing them in your production environment.

Parameter Name                            Description

----------------------------------------- ----------------------------------------------------------------

_parallel_adaptive_max_users              maximum number of users running with default DOP

_parallel_broadcast_enabled               enable broadcasting of small inputs to hash and sort merge joins

_parallel_default_max_instances           default maximum number of instances for parallel query

_parallel_execution_message_align         Alignment of PX buffers to OS page boundary

_parallel_fake_class_pct                  fake db-scheduler percent used for testing

_parallel_load_bal_unit                   number of threads to allocate per instance

_parallel_load_balancing                  parallel execution load balanced slave allocation

_parallel_min_message_pool                minimum size of shared pool memory to reserve for pq servers

_parallel_recovery_stopat                 stop at -position- to step through SMON

_parallel_server_idle_time                idle time before parallel query server dies

_parallel_server_sleep_time               sleep time between dequeue timeouts (in 1/100ths)

_parallel_txn_global                      enable parallel_txn hint with updates and deletes

_parallelism_cost_fudge_factor            set the parallelism cost fudge factor

_px_async_getgranule                      asynchronous get granule in the slave

_px_broadcast_fudge_factor                set the tq broadcasting fudge factor percentage

_px_dynamic_opt                           turn off/on restartable qerpx dynamic optimization

_px_dynamic_sample_size                   num of samples for restartable qerpx dynamic optimization

_px_granule_size                          default size of a rowid range granule (in KB)

_px_index_sampling                        parallel query sampling for index create (100000 = 100%)

_px_kxib_tracing                          turn on kxib tracing

_px_load_publish_interval                 interval at which LMON will check whether to publish PX load

_px_max_granules_per_slave                maximum number of rowid range granules to generate per slave

_px_min_granules_per_slave                minimum number of rowid range granules to generate per slave

_px_no_stealing                           prevent parallel granule stealing in shared nothing environment

_px_trace                                 px trace parameter

_PX_use_large_pool                        Use Large Pool as source of PX buffers