BW data source in-depth research [from wkingchen's blog]

Reason for forwarding: I have benefited a lot after reading it several times. It seems that I still understand the deviation. I still need to discuss how to think about before image.

 

 

Datasource is a very important part of BW. A qualified BW consultant should have a deep understanding of BW datasource.ArticleThere are also many. Let me give you a rough summary. If you have any mistakes, please correct them.

Overview of the delta mechanism for standard data sources

1. All delta data will first go to delta q and then to BW before being transmitted to BW. Delta Q can be managed and observed through rsa7. One important role of delta q is to ensure the order of records.

2. there are two cases for Delta data from the original table to delta Q: For the lo data source, the system pushes the delta data to delta Q, And then when infopackage is executed, move the data from delta Q to BW. Most non-Lo data sources use the time stamp method. During infopackage execution, the system obtains the delta data from the source data table based on the time stamp, which is sent to delta Q, then it was moved to BW. Therefore, for non-Lo data sources, it is difficult for you to see their delta data in delta Q.

3. For Delta mode of time stamp, most data sources (all ?) The time stamp of is recorded in the bwom2_timest table.

4. For the lo data source, the delta data stream varies depending on the selected _ updatemode.

 

5. How the system obtains delta data depends on the Delta process of the data source, which will be detailed later.

About Delta Process

Each data source has a very important attribute: Delta process. You can see in rsa2:

 

 

In addition, you can also find this information in the table's roosource.

Delta process tells us how the system captures delta data. Generally, delta data can be added, modified, or deleted. To add a data reserve as a delta data directly, the deletion usually generates a reserve image ). Modification is a bit complicated. There are several different processing methods:

• After image only captures the modified data as delta data.
• Add image is the difference between the captured and source data.
• Before image generates reverse data of the original data to offset the original data.

BW data source, which has the following Delta process:

 

 

How do these Delta processes capture delta data? The rodeltam table tells us the mysteries. This table records the images used by each Delta process to capture delta data and determines whether to maintain the order of records. Different image methods have a great impact on BW modeling and Data Stream:

• If before image, after image, and reserve image are used, for example, ABR (DSO's change log is also used ), then, the data can be overwritten (after image overwrites the previous before image when overwriting), and can be accumulated (When accumulating, before image writes back the original record, after image is the record after modification), which can be incrementally updated to DSO and infocube.
• If only after image is supported, data can only be overwritten and cannot be accumulated. For example, AIE can only be incrementally updated to DSO and then to infocube. In this case, there must be a DSO layer between datasource and cube.
• If only add image is supported, for example, add, the data can only be accumulated and cannot be overwritten. In this case, there is usually no DSO layer. If the DSO layer is required, set transformation to the accumulate mode.
• If it is full only, Delta must be achieved through infopackage settings.

 

I work overtime to write articles over the weekend. It's easy for me. LP is about to be used for pedicure, not yet to be continued...

About Lo datasource data streams

How to Avoid delta data loss

How can we make the datasource that only supports full implement an approximate delta update?