Content

WebSphere Data Interchange (WDI) supports three kinds of maps - Send Maps, Receive Maps, and Data Transformation (or Data Transform) Maps. The Send Maps and Receive Maps have existed with the product for a number of versions and are referred to as S/R maps. The Data Transformation maps, which are prominent in version 3.2, are referred to as DT maps.

There are two mapping tools in the WDI Client, a Send / Receive mapper and a Data Transformation mapper. When a map is created, the user designates the type of map being created. When an existing map is opened, the appropriate editor is used for the type of map being opened.

S/R maps are EDI-centric, that is, one side of the mapping must be EDI. For example, Send maps are ADF->EDI, and Receive maps are EDI->ADF. S/R maps are extremely fast and are specialized for EDI. DT maps are any-to-any based and provide more flexibility in the syntax of the source and target messages, either of which can be EDI. Record-oriented data (Data Formats/ADF/ROD), or XML. Limited tests show that DT translations use 2-4 times more CPU time than equivalent S/R maps. Some of this is related to additional language capability of DT maps - which involves internal conversion of data to Unicode. With respect to elapsed time, small (500 bytes) EDI->ADF translations using S/R technology have shown the ability to run at the rate of 100,000,000 transactions per day. DT map translations (EDI->XML) have run small EDI messages at the rate of 15,000,000 transactions per day. See the document attached for timings and test case specifics.

WDI uses a number of internal techniques during translation to increase performance.

For DT maps there are features like:

• Caching of DB objects for repeated transactions from the same trading partner
• Node pool to minimize freemain and getmain requests
• Progressive "reallocation" of output buffers to handle large messages efficiently
• Pageable AMM for handling very large messages
• Special output mechanisms for large output data

For Send / Receive maps, similar features are:

• Caching of DB objects for repeated transactions from the same trading partner
• Persistent storage usage to reduce memory requests
• Pageable Translation for large EDI messages
• Use of "virtual arrays" for internal storage of data