Commit_Every=n

    Specifies the frequency with which Oracle Rdb commits the data
    being loaded. For a single-executor load operation, Oracle Rdb
    commits the data after every n records that are stored. The
    default is to commit only after all records have been stored.

    For a parallel load operation, the Commit_Every qualifier
    applies separately to each of the executors (processes) used.
    For example, if five parallel processes are running, and the
    Commit_Every=2 qualifier is specified, Oracle RMU commits data
    for each process after it has stored 2 records. This means that
    if the Commit_Every=1000 qualifier is specified when you load one
    million records with 10 parallel processes, the .ruj files will
    store up to 10,000 rows of before-image data.

    If you specify the Defer_Index_Updates qualifier and a high value
    for the Commit_Every qualifier, memory requirements are high. See
    the description of the Defer_Index_Updates qualifier for details.
    Commit operations may occur more frequently than you specify
    under certain conditions. See the description of the Defer_Index_
    Updates qualifier for details.

    To determine how frequently you should commit data, decide how
    many records you are willing to reload if the original load
    operation fails. If you use the Statistics=On_Commit qualifier,
    you receive a message indicating the number of records loaded at
    each commit operation. Then, if a failure occurs, you know where
    to resume loading.

    If you specify the Place qualifier and a failure occurs, resume
    loading at the point of the previous commit, instead of the
    record number of the last successful commit. The Place qualifier
    restructures the .unl file prior to loading, so the record number
    on which the load operation failed does not correspond to the
    same number in the original .unl file.