Hi Tom,
We have table with 450 million rows in it. As one time activity, we copy it in lower environment and mask some of its columns (9 varchar2 columns). Masking happens through function call, it's logic cannot be changed as it is used across many heterogeneous systems. What is the best way to achieve this?

Example of UPDATE stmt:
UPDATE T
SET COL1=PKG.MASKFUNC(COL1),
COL2=PKG.MASKFUNC(COL2),
COL3=PKG.MASKFUNC(COL3),
....
Where PK BETWEEN STARTING_VALUE AND END_VALUE;
COMMIT;

This runs in loop until all rows are updated. Around 10,000 rows in single update. This took 4 days last time & want to reduce run time.

More on column data:
6 of the columns are of varchar2(400)
2 columns are of varchar2(500)
One column is varchar2(1000). Data in these columns are non-unique and may contain nulls.

Approach tried:
Mask distinct values of each column and insert into separate tables.
E.g.
INSERT /*+APPEND*/INTO MSK_COL1
WITH TEMP AS (SELECT /*+ PARALLEL(8) */ COL1 FROM T WHERE COL1 IS NOT NULL GROUP BY COL1)
SELECT COL1,MASKFUNC(COL1) AS MSKCOL1 FROM TEMP;

Similar stmts for remaining 8 columns.

CREATE UNIQUE INDEX IDX1 ON MSK_COL1 (COL1);
Exec dbms_stats.gather_table_stats(user,'T',estimate_percent=>1);
Similar stmts for remaining 8 tables.

Final insert:
INSERT /*+ APPEND */ INTO T1
SELECT /*+PARALLEL(8)*/
....
,CASE WHEN COL1 IS NOT NULL THEN
(SELECT MSKCOL1 FROM MSK_COL1 WHERE A.COL1=COL1)
ELSE NULL
END
,....similar for remaining 8 columns
...
FROM T A;

Idea here is to rename table T to T_old and T1 to T.
Tested this approach on 10 million rows, final insert itself took 72 minutes, simple extrapolation says 54 hours.
No index exists on columns being masked on table T. Table T is partitioned on Date, interval one day. T has 117 columns in it.

Please suggest.