If the latch contention is caused by TM enqueue requests associated with intensive update activity against that table, then you can resolve the problem by disabling table locks against that table. If that does not work, then your diagnosis is incorrect.

I have a Oracle 8.0.6 database in Sun Solaris 2.6 (Sun E10K, 16CPU and 6Gbytes RAM). Process running against database are Tuxedo servers (about 75). I have seen that hit ratio (misses/gets) in the latch enqueues is greater than 7%. The application servers process one input file and then they update one table (table A) in the database.

I suppose this ratio is low because Tuxedo servers are requesting a TM enqueue (level 3) in the table A of the database (hash value from this TM enqueue will be always the same, and the hash chain will be always the same). My question is how can I improve this ratio? I suppose if I increment _enqueue_hash_chain_latches I won´t get a better ratio because the hash value for the TM enqueue will be always the same (and in the same hash chain).