Bulk Insert from SQL Server to Oracle

Question

I have to load around 20 million rows from SQL Server table to Oracle table using Network Link,wrote following code using Bulk Collect,which is working but taking more time(taking 5 hrs).
I also tried with changing table to parallel degree 8 didn't help(Also Oracle Table set to NOLOGGONG mode).
Is there any better way to do this ? Appreciate any help in this regard .
Script :
CREATE OR REPLACE PROCEDURE INSERT_SQLSERVER_TO_ORACLE
IS
TYPE v_ARRAY IS TABLE OF TARGET_CUST%ROWTYPE INDEX BY BINARY_INTEGER;
ins_rows v_ARRAY;
BEGIN
DECLARE CURSOR REC1 IS
SELECT COL1, COL2,COL3,COL4 SOURCE_SQLSERVER_CUST;
BEGIN
OPEN REC1;
LOOP
FETCH REC1 BULK COLLECT INTO ins_rows LIMIT 5000;
FORALL i IN ins_rows.FIRST..ins_rows.LAST
INSERT INTO TARGET_CUST VALUES ins_rows(i);
EXIT WHEN REC1%NOTFOUND;
END LOOP;
COMMIT;
CLOSE REC1;
END;
END;
Thanks in Advance.

Answer 1

887204 wrote:
I have to load around 20 million rows from SQL Server table to Oracle table using Network Link,wrote following code using Bulk Collect,which is working but taking more time(taking 5 hrs).I would not pull that data via a network link and use standard SQL insert statements. Bulk processing is meaningless in this context. It does nothing to increase the performance - as context switching is not the issue.
The biggest factor is pulling 20 million rows's data via database link across the network. This will be slow by it's very nature.
I would use bcp (Bulk Copy export) on SQL-Server to write the data to a CSV file.
Zip that file. FTP/scp/sftp it to the Oracle server. Unzip it.
Then do a parallel direct load of the data using SQL*Loader.
This will be a lot faster than pulling uncompressed data across the network, a couple of rows at a time (together with the numerous moving parts on the Oracle side that uses a HS agent as interface between SQL-Server and the Oracle database).