Hi,

I have a data massaging logic which massages 400 million data from table and load into a target table.Both tables have same partition keys and the database is a also partitioned.My question here is which is faster cursor load or select insert with table not logged initially(NLI)? I know both doesn`t write transaction logs.I know cursor load goes through co ordinator node.How does the data move in NLI.Is it done parallely by the db2?? can someone share their knowledge on this???


Thanks,

My estimation (without trying it) is that the cursor load is faster than inserts with NLI (not logged initially), because the Load utility bypasses the DB2 data manager on the insert and lays down a page at a time (instead of processing each row as does an insert).

I don't think NLI affects anything but logging, and the insert will be the same as a regular insert in all other respects.

Thank you Marcus!! But cursor load takes the data through the co ordinator node whereas inserts use parallelism and if there are more number of CPUs doesn`t inserts are faster than cursor load?? doesn`t co ordinator node becomes a bottle neck???

Inserts have to go through coordinator node also. Load utility uses parallelism also. I would try a test if you want be sure.