BW Delta增量机制

最近比较空闲，学习了BW350的关于BW增量部分，将一些自己之前比较模糊的概念梳理了一番，顺便做个笔记；

1、Delta Process

R3中通过SE11查看表ROOSOURCE，BW中通过SE11查看表RSOLTPSOURCE可以查看数据源的抽取机制，即对应的几种抽取方式；举例如下所示：

 

ABR 采用前镜像、后镜像和反转镜像的更新模式，既支持覆盖又支持累加，所以数据源可以更新到DSO或者CUBE

AIE  采用后镜像模式，只支持覆盖，不支持累加，所以该类型数据源不能直接加载到CUBE，一般会先加载到DSO;在FI-AR/AP中此种增量处理方式应用较多

ADD 只支持累加，采用的是附加镜像的更新方式，所以既可以更新到DSO又可以更新到CUBE

 

There are various delta processes for SAP source systems (R/3, B2B, CRM...):

1. Creation of deltas with after, before and reverse images what are updated

directly to the delta queue. An after image provides the status after the

change, a before image the status before the change with a minus sign. A

reverse image sends the record with a minus sign too, indicating at the same

time that it is to be deleted. The data packages are serialized in the process.

The delta process controls whether adding or overwriting is allowed. In this

case, both adding and overwriting is allowed. This process permits data to be

updated in an ODS object as well as an InfoCube. (Technical name of the

delta process in the system: ABR)

2. The extractor provides additive deltas that are serialized on a request basis.

This serialization is necessary since the extractor provides each key once in a

request, and changes to the non-key fields would otherwise not be transferred

correctly. It only allows fields to be added. It permits data to be updated in

an ODS object as well as an InfoCube. This delta process is used by LIS

DataSources. (Technical name of the delta process in the system: ADD)

3. Creation of deltas with after image that are updated directly to the delta

queue. The data is serialized by package, as the same key can be transferred

a number of times within a request. It does not allow data to be updated

directly to an InfoCube. To update an InfoCube, you always have to have an

ODS object as a go-between. With numeric key figures, for example, this

process only allows you to overwrite data. Adding is not permitted, as this

would produce incorrect results. It is used to transfer line items in FI-AP/AR,

whilst the variation on this process, in which the extractor can also send

records marked for deletion, is used in this function in the BBP. (Technical

name of the delta process in the system: AIM/AIMD)

 

其中在ROOSOUCE中，还存在着提取方式字段，即下图红色圈图所示

 

有三种提取方式，分别是F1(提取结构)、V（VIEW OR TABLE）和 Q(INFOSET),分别对应到创建一般数据源时候可以采用的三种方式

此外，在后一字段中，还可以看到每个数据源的提取结构；

 

上图的ZDD_ABLE字段表示是否支持“早期增量初始化”，这个表示在初始化时候是否继续收集增量；

提取器确定在初始化请求时是否降数据写入增量队列或写入应用程序的增量表中。

在运行初始化请求时，若在源系统中用该数据请求调用的提取器可以写入增量记录到增量列队，则可以使用该标识控制是否执行初期增量初始化。

If you carry out an Early Delta Initialiation, the process is as follows:

1. The selection for the delta initialization of the DataSource is saved and the

delta queue for the DataSource is generated in the source system.

2. All the data that corresponds to the selection criteria determined in the

InfoPackage is requested (similar to the full update).

The steps ìGenerate Delta Queue? and Request Data are reversed here

comparison with the normal way in hich the delta process is initialized!

Use the Early Delta Initialization indicator in the InfoPackage if you want t

avoid a period when no updates can be made (no document processing and update)

in the source system while the delta process is being initialized.

Note that in some applications (such as LO Cockpit), an update-free period is

nevertheless needed to fill the setup tables. In such cases, the update-free time

is merely reduced.

If you want to carry out early delta initialization, the update in the source

system can continue (A) and the data can be written to the delta queue while the

initialization requests are being processed.

 

Update Method

 

什么叫V3 Update Mode?

V1 同步更新模式，即凭证产生就更新增量，与业务数据同步更新；

V2 异步更新模式，就如一个两步的操作一样，第一步业务凭证更新了，然后再更新第二步的数据源增量表

V3 异步更新模式，与V2的区别在于他的更新时通过后台事件来触发的，即定一个任务定是收集增量并更新至增量表

 

知道了3种更新模式后，再来看LBWE中的几种增量模式

Direct Delta:这就是一种V1模式，数据同步更新到增量队列，这种模式系统负荷很重，特别是对于业务量大的凭证；

Queued Delta：类似于V3的更新模式，与V3更新的区别在于，数据先被收集到一个抽取队列中（V1模式），然后被送到增量队列（V3模式）；见下图

Unsterilized V3 Update：此类型的最大特征是没有序列化，就是说增量队列中凭证是无序的，这个对于采用覆盖模式的模型来说是最致命的，所以如果目标队列是DSO的话，还是不要采用这种模式好

 

LUW

LUW = Logic Unit of Work,为理论上的工作单元，就是一个系统设置的理论包大小

通过RSA7可以查看增量队列如下

 

其中的TOTAL不是该增量队列的数据，而是包的数据；

 

DELTA UPDATE & DELTA REPETION

其中Delta Update表示当前增量队列的数据

Delta Repetition表示的是上一次提取的增量数据

 

数据源的支持的Image详解

根据每个数据源的更新模式，可以具体了解其支持的Image模式，也可以得到数据目标应该采用哪种更新方式（DSO OR CUBE 累计还是覆盖）

SE11进入表RODELTAM，查看其中每种类型支持的Image如下

 

 

例如对比AIE,由于0FI_GL_4即是采用AIE 增量处理模式，所以其职能支持后印象，也就是说只能采用覆盖，不能采用累计的方式