i have little bit confusion regarding sequential and parallel processing ,when to go for sequential and parallel processing,what is the advantage ?
please help me,i am new to odi.
The advantage ? In simple terms, say you have 3 steps that take 10 minutes, in serial :
Step 1 -> Step 2 -> step 3
10 + 10 + 10 = 30 Minutes to complete.
Now lets say you can run these steps in Parallel at the same time
They will all be finished in 10 minutes, you have saved 20 minutes running in Parallel compared to Serial. In a data warehouse, we can commonly load Dimensions in parallel (providing there are no dependancies) , then we might load the fact tables in parallel also (again , assuming no dependancies).
Load Plans allow you to orchestrate Parallel / Serial calls to interfaces and Package scenarios, you can also run steps in Parallel in Packages using asynchronous scenario calls.
Hope this helps.
The types of package depend on what type of integration you are trying to perform.
In 10G - Packages where used to create a process flow of interfaces and ODITools required to perform whatever task you wanted to achieve, with success and fail flow defined for error handling.
In 11G I see clients moving away from packages more and just using a Load Plan to launch the interface scenarios.
Packages do expose all the ODItools (Change Data capture, FTP, email etc.) in a graphical UI and allow you to specifcy Pass and Fail actions on a given step - Remember that you can call an ODITool from an ODI Procedure though as well though.
As for Incremental / Full load - these are generic ETL loading terms, Full Load signifies that you will load all source data and integrate into the DWH, Incremental load suggests you will only be loading data that has changed since the last run, you might use a Date Modified or some form of Change Data Capture to determine what has changed.