hi,biti_rainy; I don't think parallel process's locks are all put on the same list. If so,Does it mean all parallel will fall into contention destined? And we have _enqueue_hash_chain_latches =6 ,default to cpu_count. With parallel 4 and _enqueue_hash_chain_latches =6, and why so much contention? Maybe the explain is just like you said :the parallel process's locks are all put on the same list. On Sun, 9 Jan 2005 12:14:42 +0800, biti_rainy <biti_rainy@xxxxxxxxx> wrote: > hi,eygle > > i think the 'enqueue hash chains' looks like the 'cache buffer > chains',oracle often use * hash table + list * to manage resource. > enqueue resources would be put on the list blow some hash bucket. > so ,i am not sure whether the parallel process's locks are all put on > the same list. if they always on the same list ,you would not use > parallel insert. if not,you would increase the _enqueue_hash to a prime > number. > > there is my 9i database: > sys@OCN>select INDX from x$ksppi where KSPPINM = '_enqueue_hash'; > > INDX > ---------- > 82 > > sys@OCN>select KSPPSTVL from X$KSPPSV where indx = 82; > > KSPPSTVL > ---------------------------------------------------------------------------------------------------- > 177 > > sys@OCN>show parameter dml_locks > > NAME TYPE VALUE > ------------------------------------ ----------- > ------------------------------ > dml_locks integer 312 > sys@OCN>show parameters enqueue_resource > > NAME TYPE VALUE > ------------------------------------ ----------- > ------------------------------ > enqueue_resources integer 532 > sys@OCN> > > > Best regards > > msn: biti_rainy@xxxxxxxxxxx > a dba from alibaba(china) > ---------------------------------------- eygle,a beginer of Oracle from China. my site: http://www.eygle.com -- //www.freelists.org/webpage/oracle-l