Monday, September 28, 2026

Hidden Parameters in Oracle AI Database 26ai H.9 Parallel execution (171)

 

NameTypeValueAt default?LevelPDBRAC inst.FlagsDescription
_parallel_adaptive_max_usersInteger4yesSYSTEM imm.NYmaximum number of users running with default DOP
_parallel_blackbox_enabledBooleanTRUEyesSESSIONYYparallel execution blackbox enabled
_parallel_blackbox_sgaBooleanTRUEyesSYSTEM imm.NYtrue if blackbox will be allocated in SGA, false if PGA
_parallel_blackbox_sizeInteger16384yesSESSIONYYparallel execution blackbox bucket size
_parallel_broadcast_enabledBooleanTRUEyesSESSIONYYenable broadcasting of small inputs to hash and sort merge joins
_parallel_cluster_cache_pctInteger80yesSESSIONYYmax percentage of the global buffer cache to use for affinity
_parallel_cluster_cache_policyStringADAPTIVEyesSESSIONYYpolicy used for parallel execution on cluster(ADAPTIVE/CACHED)
_parallel_conservative_queuingBooleanFALSEyesSYSTEM imm.NYconservative parallel statement queuing
_parallel_ctas_enabledBooleanTRUEyesSESSIONYYenable/disable parallel CTAS operation
_parallel_default_max_instancesInteger1yesSTATICNNdefault maximum number of instances for parallel query
_parallel_execution_message_alignBooleanFALSEyesSTATICNNAlignment of PX buffers to OS page boundary
_parallel_fake_class_pctInteger0yesSESSIONYYfake db-scheduler percent used for testing
_parallel_fault_tolerance_enabledBooleanFALSEyesSTATICNNenables or disables fault-tolerance for parallel statement
_parallel_fault_tolerance_thresholdInteger3yesSESSIONYYtotal number of faults fault-tolerance will handle
_parallel_fixwrite_bucketInteger1000yesSTATICNNNumber of buckets for each round of fix write
_parallel_frz_step_reconfigInteger0yesSYSTEM imm.NYparallel rs work for freeze step for instance reconfig.
_parallel_heartbeat_snapshot_intervalInteger2yesSYSTEM imm.NYinterval of snapshot to track px msging between instances
_parallel_heartbeat_snapshot_maxInteger128yesSYSTEM imm.NYmaximum number of historical snapshots archived
_parallel_inmemory_min_time_thresholdStringAUTOyesSESSIONYYthreshold above which a plan is a candidate for parallelization for in-memory tables (in seconds)
_parallel_inmemory_time_unitInteger1yesSESSIONYYunit of work used to derive the degree of parallelism for in-memory tables (in seconds)
_parallel_lmd_reconfigInteger1yesSYSTEM imm.NYparallel lmd work in reconfiguration for enqueues.
_parallel_load_bal_unitInteger0yesSESSIONYYnumber of threads to allocate per instance
_parallel_load_balancingBooleanTRUEyesSESSIONYYparallel execution load balanced slave allocation
_parallel_load_publish_thresholdInteger10yesSESSIONYYdiffrence in percentage controlling px load propagation
_parallel_log_write_cleanupBooleanTRUEyesSTATICNNPerform (null) log write cleanup in parallel
_parallel_min_message_poolString491520yesSTATICNNminimum size of shared pool memory to reserve for pq servers
_parallel_optimization_phase_for_localBooleanFALSEyesSESSIONYYparallel optimization phase when all slaves are local
_parallel_queuing_max_waitingtimeIntegeryesSYSTEM imm.NYparallel statement queuing: max waiting time in queue
_parallel_recovery_stopatInteger32767yesSYSTEM imm.NYstop at -position- to step through SMON
_parallel_replay_msg_limitInteger4000yesSTATICNNNumber of messages for each round of parallel replay
_parallel_scalabilityInteger50yesSESSIONYYParallel scalability criterion for parallel execution
_parallel_server_idle_timeInteger30000yesSYSTEM imm.NYidle time before parallel query server dies (in 1/100 sec)
_parallel_server_memory_thresholdBig integer0yesSYSTEM imm.NYmax memory threshold for idle PX server to exit
_parallel_server_resiliencyInteger0yesSESSIONYYenable resilient execution of parallel query
_parallel_server_sleep_timeInteger1yesSYSTEM imm.NYsleep time between dequeue timeouts (in 1/100ths)
_parallel_slave_acquisition_waitInteger1yesSESSIONYYtime(in seconds) to wait before retrying slave acquisition
_parallel_statement_queuingBooleanFALSEyesSESSIONYYparallel statement queuing enabled
_parallel_syspls_obey_forceBooleanTRUEyesSESSIONYYTRUE to obey force parallel query/dml/ddl under System PL/SQL
_parallel_time_unitInteger10yesSESSIONYYunit of work used to derive the degree of parallelism (in seconds)
_parallel_txn_globalBooleanFALSEyesSESSIONNNenable parallel_txn hint with updates and deletes
_parallelism_cost_fudge_factorInteger350yesSESSIONYYset the parallelism cost fudge factor
_px_adaptive_dist_bypass_enabledBooleanTRUEyesSESSIONYYenable the bypass table queue for adaptive distribution methods
_px_adaptive_dist_bypass_optimizationString1yesSESSIONYYsets optimizations for bypass table queue for ADM
_px_adaptive_dist_methodStringCHOOSEyesSESSIONYYdetermines the behavior of adaptive distribution methods
_px_adaptive_dist_method_thresholdInteger0yesSESSIONYYBuffering / decision threshold for adaptive distribution methods
_px_adaptive_dist_nij_enabledStringONyesSESSIONYYenable adaptive distribution methods for left non-inner joins
_px_adaptive_offload_percentageInteger30yesSESSIONYYpercentage for PQ adaptive offloading of granules
_px_adaptive_offload_thresholdInteger10yesSESSIONYYthreshold (GB/s) for PQ adaptive offloading of granules
_px_autodop_pq_overheadBooleanTRUEyesSESSIONYYadjust auto dop calculation using pq overhead
_px_back_to_parallelStringOFFyesSESSIONYYallow going back to parallel after a serial operation
_px_bind_peek_sharingBooleanTRUEyesSESSIONYYenables sharing of px cursors that were built using bind peeking
_px_broadcast_fudge_factorInteger100yesSESSIONYYset the tq broadcasting fudge factor percentage
_px_buf_subheap_sizeInteger4096yesSTATICNNbuffer subheap size
_px_buffer_ttlInteger30yesSESSIONYYttl for px mesg buffers in seconds
_px_cdb_view_enabledBooleanTRUEyesSESSIONYYparallel cdb view evaluation enabled
_px_cdb_view_join_enabledBooleanTRUEyesSESSIONYYdisable parallelism cap on CDB view
_px_chunk_sizeInteger1000000yesSESSIONYYdefault size of a chunk for row id granules (in KB)
_px_chunklist_count_ratioInteger32yesSTATICNNratio of the number of chunk lists to the default DOP per instance
_px_compilation_debugInteger0yesSESSIONYYdebug level for parallel compilation
_px_compilation_traceInteger0yesSESSIONYNtracing level for parallel compilation
_px_concurrentBooleanTRUEyesSESSIONYYenables pq with concurrent execution of serial inputs
_px_cpu_autodop_enabledBooleanTRUEyesSESSIONYYenables or disables auto dop cpu computation
_px_cpu_operator_bandwidthStringyesSESSIONYYCPU operator bandwidth in MB/sec for DOP computation
_px_cpu_process_bandwidthInteger200yesSESSIONYYCPU process bandwidth in MB/sec for DOP computation
_px_dist_agg_partial_rollup_pushdownStringADAPTIVEyesSESSIONYYperform distinct agg partial rollup pushdown for px execution
_px_dp_array_sizeInteger32767yesSTATICNNMax number of pq processes supported
_px_dump_12805_sourceBooleanTRUEyesSESSIONYYenables or disables tracing of 12805 signal source
_px_dup_proc_checkBooleanFALSEyesSYSTEM imm.NYenable duplicate parallel process check with name service
_px_dynamic_granulesBooleanTRUEyesSESSIONYYenable dynamic granule generation
_px_dynamic_granules_adjustInteger10yesSESSIONYYadjust dynamic granule regeneration
_px_dynamic_optBooleanTRUEyesSESSIONYYturn off/on restartable qerpx dynamic optimization
_px_dynamic_sample_sizeInteger50yesSESSIONYYnum of samples for restartable qerpx dynamic optimization
_px_execution_services_enabledBooleanTRUEyesSESSIONYYenable service-based constraint of px slave allocation
_px_extended_join_skew_handlingBooleanTRUEyesSESSIONYYenables extended skew handling for parallel joins
_px_external_table_default_statsBooleanTRUEyesSESSIONYYthe external table default stats collection enable/disable
_px_filter_parallelizedBooleanTRUEyesSESSIONYYenables or disables correlated filter parallelization
_px_filter_skew_handlingBooleanTRUEyesSESSIONYYenable correlated filter parallelization to handle skew
_px_flood_ctrl_timeInteger900yesSESSIONYYparallel flood control time in seconds
_px_freelist_latch_divisorInteger2yesSTATICNNDivide the computed number of freelists by this power of 2
_px_gim_factorInteger100yesSESSIONYYweighted autodop global index maintenance factor
_px_granule_alignmentInteger1024yesSESSIONYYdefault size of a rowid range granule (in KB)
_px_granule_batch_sizeInteger26yesSESSIONYYmaximum size of a batch of granules
_px_granule_randomizeBooleanTRUEyesSESSIONYYenables or disables randomization of parallel scans rowid granules
_px_granule_sizeInteger1000000yesSESSIONYYdefault size of a rowid range granule (in KB)
_px_groupby_pushdownStringFORCEyesSESSIONYYperform group-by pushdown for parallel query
_px_hold_timeInteger0yesSESSIONYYhold px at execution time (unit: second)
_px_hybrid_TSM_HWMB_loadBooleanTRUEyesSESSIONYYEnable Hybrid Temp Segment Merge/High Water Mark Brokered load method
_px_hybrid_partition_execution_enabledBooleanTRUEyesSESSIONYYenable parallel hybrid partition execution
_px_hybrid_partition_skew_thresholdInteger10yesSESSIONYYpartitions bigger than threshold times average size are skewed
_px_increase_join_frag_sizeBooleanTRUEyesSYSTEM imm.NYincrease join message fragment size
_px_index_sampling_objsizeBooleanTRUEyesSESSIONYYparallel query sampling for index create based on object size
_px_io_process_bandwidthInteger200yesSESSIONYYIO process bandwidth in MB/sec for computing DOP
_px_io_system_bandwidthInteger0yesSESSIONYYtotal IO system bandwidth in MB/sec for computing DOP
_px_join_skew_handlingBooleanTRUEyesSESSIONYYenables skew handling for parallel joins
_px_join_skew_minfreqInteger30yesSESSIONYYsets minimum frequency(%) for skewed value for parallel joins
_px_join_skew_null_handlingBooleanTRUEyesSESSIONYYenables null skew handling improvement for parallel outer joins
_px_join_skew_ratioInteger10yesSESSIONYYsets skew ratio for parallel joins
_px_join_skew_sampling_percentInteger0yesSESSIONYYSets sampling percentage for skew handling sampling queries
_px_join_skew_sampling_time_limitInteger50yesSESSIONYYSets execution time limit for skew handling sampling queries
_px_join_skew_use_histogramBooleanTRUEyesSESSIONYYEnables retrieval of skewed values from histogram when possible
_px_join_skewed_values_countInteger0yesSESSIONYYforce join skew handling with specified skewed values count
_px_kxib_tracingInteger0yesSESSIONYYturn on kxib tracing
_px_load_balancing_policyStringUNIFORMyesSESSIONYYparallel load balancing policy
_px_load_factorInteger300yesSESSIONYYweighted autodop load factor
_px_load_monitor_thresholdInteger10000yesSESSIONYYthreshold for pushing information to load slave workload monitor
_px_load_publish_intervalInteger200yesSESSIONYYinterval at which LMON will check whether to publish PX load
_px_loc_msg_costInteger1000yesSESSIONYYCPU cost to send a PX message via shared memory
_px_max_granules_per_slaveInteger100yesSESSIONYYmaximum number of rowid range granules to generate per slave
_px_max_map_valInteger32yesSYSTEM imm.NYMaximum value of rehash mapping for PX
_px_max_message_pool_pctInteger40yesSTATICNNpercentage of shared pool for px msg buffers range [5,90]
_px_message_compressionBooleanTRUEyesSESSIONYYenable compression of control messages for parallel query
_px_min_granules_per_slaveInteger13yesSESSIONYYminimum number of rowid range granules to generate per slave
_px_min_ndv_per_slave_perc_funcInteger2yesSESSIONYYmininum NDV of TQ keys needed per slave for parallel percentile
_px_minus_intersectBooleanTRUEyesSESSIONYYenables pq for minus/interect operators
_px_net_msg_costInteger10000yesSESSIONYYCPU cost to send a PX message over the internconnect
_px_nlj_bcast_rr_thresholdInteger10yesSESSIONYYthreshold to use broadcast left random right distribution for NLJ
_px_no_granule_sortBooleanFALSEyesSESSIONYYprevent parallel partition granules to be sorted on size
_px_no_stealingBooleanFALSEyesSESSIONYYprevent parallel granule stealing in shared nothing environment
_px_nss_planbBooleanTRUEyesSESSIONYYenables or disables NSS Plan B reparse with outline
_px_numa_stealing_enabledBooleanTRUEyesSESSIONYYenable/disable PQ granule stealing across NUMA nodes
_px_numa_support_enabledBooleanFALSEyesSYSTEM imm.NYenable/disable PQ NUMA support
_px_object_samplingInteger200yesSESSIONYYparallel query sampling for base objects (100000 = 100%)
_px_object_sampling_enabledBooleanTRUEyesSESSIONYYuse base object sampling when possible for range distribution
_px_object_sampling_multiplierInteger2yesSESSIONYYnumber of samples to send to QC per slave is DOP * prmixsm
_px_onepass_slave_acquisitionBooleanTRUEyesSESSIONYYenable/disable one pass slave acquisition for parallel execution
_px_oneslave_per_instance_parseBooleanTRUEyesSESSIONYYsend parse to one slave per instance
_px_overhead_exec_ctrlmesgInteger4yesSESSIONYYpq overhead on sending one control message during execution (in microseconds)
_px_overhead_init_endpointsInteger4yesSESSIONYYpq overhead on setting up one end point during initialization (in microseconds)
_px_overhead_init_slaveallocInteger1000yesSESSIONYYpq overhead on allocating per slave during initialization (in microseconds)
_px_overhead_teardownInteger600yesSESSIONYYpq overhead on setting up one end point during initialization (in microseconds)
_px_parallel_spawn_min_countInteger2yesSTATICNNPQ parallel spawn min threshold count
_px_parallelize_expressionBooleanTRUEyesSESSIONYYenables or disables expression evaluation parallelization
_px_parallelize_non_native_datatypeBooleanTRUEyesSESSIONYYenable parallelization for non-native datatypes
_px_partial_rollup_pushdownStringADAPTIVEyesSESSIONYYperform partial rollup pushdown for parallel execution
_px_partition_load_dist_thresholdInteger64yesSESSIONYY# of load slave to # of partition ratio to switch distribution
_px_partition_load_skew_handlingStringOFFyesSESSIONYYenable partition load skew handling
_px_partition_load_skew_thresholdInteger3yesSESSIONYYpartition loads bigger than threshold times average are skewed
_px_partition_scan_enabledBooleanTRUEyesSESSIONYYenables or disables parallel partition-based scan
_px_partition_scan_thresholdInteger64yesSESSIONYYleast number of partitions per slave to start partition-based scan
_px_partition_skew_thresholdInteger80yesSESSIONYYpercentage of table size considered as threshold for skew determination
_px_proactive_slave_alloc_thresholdInteger8yesSESSIONYYparallel proactive slave allocation threshold/unit
_px_proactive_slave_releaseInteger1yesSESSIONYYearly release of producer slaves of final DFO
_px_proc_constrainBooleanTRUEyesSYSTEM imm.NYreduce parallel_max_servers if greater than (processes - fudge)
_px_pwg_enabledBooleanTRUEyesSESSIONYYparallel partition wise group by enabled
_px_pwise_wif_enabledBooleanTRUEyesSESSIONYYenable parallel partition wise window function
_px_pwmr_enabledBooleanTRUEyesSESSIONYYparallel partition wise match recognize enabled
_px_replication_enabledBooleanTRUEyesSESSIONYYenables or disables replication of small table scans
_px_reuse_server_groupsStringMULTIyesSESSIONYYenable/disable reusing of server groups that are already acquired
_px_round_robin_rowcntInteger1000yesSESSIONYYround robin row count to enq to next slave
_px_rownum_pdBooleanTRUEyesSESSIONYYturn off/on parallel rownum pushdown optimization
_px_scalable_gby_invdistBooleanTRUEyesSESSIONYYscalable PX plan for GBY w/ inverse distribution functions
_px_scalable_invdistBooleanTRUEyesSESSIONYYenable/disable px scalable plan for inverse distribution functions
_px_scalable_invdist_mcolBooleanTRUEyesSESSIONYYenable/disable px plan for percentile functions on multiple columns
_px_send_timeoutInteger300yesSESSIONYYIPC message send timeout value in seconds
_px_shared_hash_joinBooleanFALSEyesSESSIONYYenable/disable shared hash join
_px_single_server_enabledBooleanTRUEyesSESSIONYYallow single-slave dfo in parallel query
_px_slave_parse_first_with_outline_hintsBooleanTRUEyesSESSIONYYEnable/Disable trying outlined plan first
_px_split_batches_per_slaveInteger7yesSESSIONYYnumber of split granules per slave for exadoop
_px_split_multi_msgBooleanTRUEyesSESSIONYYuse multiple messages to send batch of split granules
_px_split_use_single_listBooleanTRUEyesSESSIONYYuse single list for split granules
_px_tq_memcpy_thresholdInteger100yesSESSIONYYthreshold for PQ TQ to use memcpy for packing columns
_px_tq_rowhvsBooleanTRUEyesSESSIONYYturn on intra-row hash valueing sharing in TQ
_px_traceStringnoneyesSESSIONNNpx trace parameter
_px_ual_serial_inputBooleanTRUEyesSESSIONYYenables new pq for UNION operators
_px_wif_dfo_declumpingStringCHOOSEyesSESSIONYYNDV-aware DFO clumping of multiple window sorts
_px_wif_extend_distribution_keysBooleanTRUEyesSESSIONYYextend TQ data redistribution keys for window functions
_px_wif_min_ndv_per_slaveInteger2yesSESSIONYYmininum NDV of TQ keys needed per slave for scalable WiF PX
_px_window_skew_handlingBooleanTRUEyesSESSIONYYenable window function skew handling
_px_window_skew_sample_sizeInteger1000yesSESSIONYYwindow function skew sample size
_px_xt_granule_sampling_ratioInteger200yesSESSIONYYrange table queue object sampling ratio for external table
_px_xtgranule_sizeInteger10000yesSESSIONYYdefault size of a external table granule (in KB)

No comments:

Post a Comment