What is difference between PSA and ALE IDoc?  And how data is transferd using each one of them?

The following update types are available in SAP BW:

1. PSA

2. ALE (data IDoc) 

You determine the PSA or IDoc transfer method in the transfer rule maintenance screen. The process for loading the data for both transfer methods is triggered by a request IDoc to the source system. Info IDocs are used in both transfer methods. Info IDocs are transferred exclusively using ALE

A data IDoc consists of a control record, a data record, and a status record The control record contains, for example, administrative information such as the receiver, the sender, and the client. The status record describes the status of the IDoc, for example, "Processed".  If you use the PSA for data extraction, you benefit from increased flexiblity (treatment of incorrect data records). Since you are storing the data temporarily in the PSA before updating it in to the data targets, you can check the data and change it if necessary. Unlike a data request with IDocs, the PSA gives you various options for additional data updates into data targets:

InfoObject/Data Target Only - This option means that the PSA is not used as a temporary store. You choose this update type if you do not want to check the source system data for consistency and accuracy, or you have already checked this yourself and are sure that you no longer require this data since you are not going to change the structure of the data target again.

PSA and InfoObject/Data Target in Parallel (Package by Package) - BW receives the data from the source system, writes the data to the PSA and at the same time starts the update into the relevant data targets.  Therefore, this method has the best performance.

The parallel update is described in detail in the following: A dialog process is started by   data package, in which the data of this package is writtein into the PSA table. If the data is posted successfully into the PSA table, the system releases a second, parallel dialog process that writes the data to the data targets. In this dialog process the transfer rules for the data records of the data package are applied, that data is transferred to the communcation structure, and then written to the data targets. The first dialog process (data posting into the PSA) confirms in the source system that is it completed and the source system sends a new data package to BW while the second dialog process is still updating the data into the data targets.

The parallelism relates to the data packages, that is, the system writes the data packages into the PSA table and into the data targets in parallel.  Caution: The maximum number of processes setin the source system in customizing for the extractors does not restrict the number of processes in BW. Therefore, BW can require many dialog processes for the load process. Ensure that there are enough dialog processes available in the BW system. If there are not enough processes on the system side, errors occur. Therefore, this method is the least recommended. 

PSA and then into InfoObject/Data Targets (Package by Package) - Updates data in series into the PSA table and into the data targets by data package. The system starts one process that writes the data packages into the PSA table. Once the data is posted successfuly into the PSA table, it is then written to the data targets in the same dialog process. Updating in series gives you more control over the overall data flow when compared to parallel data transfer since there is only one process per data package in BW. In the BW system the maximum number of dialog process required for each data request corresponds to the setting that you made in customizing for the extractors in the control parameter maintenance screen. In contrast to the parallel update, the system confirms that the process is completed only after the data has been updated into the PSA and also into the data targets for the first data package.

Only PSA - The data is not posted further from the PSA table immediately. It is useful to transfer the data only into the PSA table if you want to check its accuracy and consistency and, if necessary, modify the data. You then have the following options for updating data from the PSA table:

Automatic update - In order to update the data automatically in the relevant data target after all data packages are in the PSA table and updated successfully there, in the scheduler when you schedule the InfoPackage, choose Update Subsequently in Data Targets on the Processing tab page.   *-- Sunil

What is difference between PSA and ODS?

PSA: This is just an intermediate data container. This is NOT a data target. Main purpose/use is for data quality maintenance. This has the original data (unchanged) data from source system.

ODS: This is a data target. Reporting can be done through ODS. ODS data is overwriteable. For datasources for which delta is not enabled, ODS can be used to upload delta records to Infocube.  

You can do reporting in ODS. In PSA you can't do reporting directly

ODS contains detail -level data , PSA The requested data is saved, unchanged from the source system. Request data is stored in the transfer structure format in transparent, relational database tables in the Business Information Warehouse. The data format remains unchanged, meaning that no summarization or transformations take place

In ODS you have 3 tables Active, New data table, change log, In PSA you don't have.