COMMIT statements are used to release locks which are required for that unit of work, and then permit a new unit of work. In other words, if COMMITS are not coded in the program and the program has been sent for execution, then while processing, in place of just going back to a few inserts since the nearest commit, the program has to go back to the inserts which were made during the entire run of the program. This extra process takes around twice or thrice the time taken normally.