Loading data from SQL server to Oracle database

Question

i want to create one table in oracle db from table in sql server. Table is huge it got 97,456,789 records.
I created db link (HS) in oracle database which is pointing to sql server. i can select that table from oracle thru db link.
select * from "dbo"."t1@dblink;
i fired below create table.
create table t2 nologging parallel (degree=3) as select * from "dbo"."t1@dblink;
and its taking long time.... but its running...
is there any alternate method to do this and and populate table in oracle db faster.
Please advise. thanks.

Answer 1

vhiware wrote:
create table t2 nologging parallel (degree=3) as select * from "dbo"."t1@dblink;
and its taking long time.... but its running...I doubt that parallel processing will be used as this is unique to Oracle (using rowid ranges typically) and not SQL-Server.
is there any alternate method to do this and and populate table in oracle db faster.Part of the performance overhead is pulling that data from SQL-Server to Oracle across the network link between them. This can be sped up by compressing the data first - and then transferring that across the network.
For example: using +bcp+ to export the data on the SQL-Server box to a CSV file, compress/zip the file, scp/sftp the file to Oracle, and then unzipping it there. Parallel and direct load processing can now be done using SQL*Loader to load the CSV into Oracle.
If this is a Linux/Unix base system, then the unzip/decompress process can be run in parallel with the SQL*Loader process by creating a pipe between the two - where the unzip process writes uncompressed data into the pipe and SQL*Loader reads and loads data as it becomes available via the pipe.
Alternatively, you can do you own PQ processing. Let's say the data is date ranged. You can create a procedure on Oracle that looks something like this:
{code}
create or replace procedure CopyDay( day date ) is
begin
insert /*+ append */ into local_tab select * from remote_tab@remotedb where col_day = day;
-- add logging info, commit, etc.
end;
{code}
You can now start 10 or more of these for different days and run it in the background using DBMS_JOB.