Best way to insert millions of records into the table

Question

Hi,
Performance point of view, I am looking for the suggestion to choose best way to insert millions of records into the table.
Also guide me How to implement in easier way to make better performance.
Thanks,
Orahar.

Answer 1

Orahar wrote:
Its Distributed data. No. of clients and N no. of Transaction data fetching from the database based on the different conditions and insert into another transaction table which is like batch process.Sounds contradictory.
If the source data is already in the database, it is centralised.
In that case you ideally do not want the overhead of shipping that data to a client, the client processing it, and the client shipping the results back to the database to be stored (inserted).
It is must faster and more scalable for the client to instruct the database (via a stored proc or package) what to do, and that code (running on the database) to process the data.
For a stored proc, the same principle applies. It is faster for it to instruct the SQL engine what to do (via an INSERT..SELECT statement), then pulling the data from the SQL engine using a cursor fetch loop, and then pushing that data again to the SQL engine using an insert statement.
An INSERT..SELECT can also be done as a direct path insert. This introduces some limitations, but is faster than a normal insert.
If the data processing is too complex for an INSERT..SELECT, then pulling the data into PL/SQL, processing it there, and pushing it back into the database is the next best option. This should be done using bulk processing though in order to optimise the data transfer process between the PL/SQL and SQL engines.
Other performance considerations are the constraints on the insert table, the triggers, the indexes and so on. Make sure that data integrity is guaranteed (e.g. via PKs and FKs), and optimal (e.g. FKs should be indexes on the referenced table). Using triggers - well, that may not be the best approach (like for exampling using a trigger to assign a sequence value when it can be faster done in the insert SQL itself). Personally, I avoid using triggers - I rather have that code residing in a PL/SQL API for manipulating data in that table.
The type of table also plays a role. Make sure that the decision about the table structure, hashed, indexed, partitioned, etc, is the optimal one for the data structure that is to reside in that table.