Problems inserting huge ammount of lines in a table

Question

Dear all:
I have a process that reads data from one table (table A), makes some calculations and writes the modified data into another table (table B).
For each line on table A the process will write from 1 to 4 lines in table B. The calculation is VERY simple.
The process reads 1 million lines from table A and writes 4 million lines into table B, and I need to do this in 2 hours.
I have a sun/solaris server with 8 processors and an EMC storage (the machine is really fast), but I can not complete the process in a reasonable time.
The problem is that after 15 minutes the Oracle Database slows A LOT. If I query the v$session_wait view, I receive "log buffer wait" messages, and there is nothing I can do to solve this problem (many DBAs tried to solve it).
Is there a special way to insert a huge amount of data into a table in a fast way?
I really appreciate any help!

Answer 1

I am extracting once from the source and inserting into the destinations table making calculations from the cursor variables.
I heard that the problem is a big contention at the Oracle i/o slave. I will try to use direct load and paralellize the process, I hope this can work!
Thank you!
Josi,
Since you are extracting from one source table and inserting into a target table, and you mentioned your are generating for each source 4 rows into the target, i like know knwo what your OWB mapping look like; i.e. how did you model the "row-multiplier" in the OWB mapping editor.... using a splitter like this:
+----> Calculate()-->Target1
+
Source1 ---> Splitter +----> Calculate()--> Target1
+
+----> Calculate()--> Target1
+
+----> Calculate()--> Target1
If you use the splitter and run the map in set-based mode, you are effectively extracting the table Source1 four times... quite costly... you probably get better performance by running the map in row-based mode and using the bulk mode..... again my recommendations are based on guesses what your mapping looks like and where/when you do the calculation in the dataflow graph.
In 9i there is a multi table insert statement which will provide you with a set-based splitter operation; we will make support available for multi table insert in our upcoming release.