Loading Data on Multiple Threads

Large data files, such as sales and inventory, can support multiple threads for parallel execution, which can greatly reduce the load times for history files. Some tables have an explicit column called ETL_THREAD_VAL to define your threads for incoming data. Parameters are available in C_ODI_PARAM_VW (in the Control Center) to configure each load program's maximum number of threads, using the parameter name LOC_NUM_OF_THREAD. For larger retailers, it is common to define 5 to 8 threads for incoming data feeds. You may choose how you split data across threads when generating the files, but a common strategy is to assign stores to threads, such that all the data for a given location is processed on the same thread.

For many interfaces, the ETL_THREAD_VAL column has been removed and, in those cases, the system will automatically calculate the threads based on the LOC_NUM_OF_THREAD parameter values. If ETL_THREAD_VAL is not provided on a file that has the column, the same automation will be applied to determine multi-threading behavior. The LOC_NUM_OF_THREAD values must be updated appropriately by the implementer in C_ODI_PARAM to leverage multi-threading capabilities.