Hi, Can anyone please let me know if I load a file using Teradata Parallel transport 12 , using multiple read and load instances, the order of records will be maintained or not ?
Example : Input file has 10 records. Will the 10 records be processed sequentially for upsert , as mentioned in the file ?
' UPDATE t1 SET WHERE .. ' , 'INSERT INTO t1 VALUES .. ') ,
INSERT For MISSING Update ROWS
TO OPERATOR ( load_op )
from OPERATOR ( RO_INPUTLAYOUT );
Since the file would contain records , inorder of update date, we need to perform upsert sequentially, else it might impact data.
A few things:
1. TPT is a parallel application. By nature, parallel applications do not rely on data order.
2. You cannot use the Load operator to perform UPSERTs. The Load operator can only insert into empty tables.
3. For UPSERTs you have to use the Update or Stream operator.
4. If you want to guarantee order, use a single instance of the operators.
Appreciate your quick reply.
I have an mload job using standalone Multiload utility that upserts around 240 million records from file . It takes 5 hrs to complete this , using 8 sessions.
Instead of Mload, if I use TPT with update operator and single read & load instances (to ensure order ) help improve the performance?
Please suggest how we can improve performance in this case ?