Relationships link entities together, so to run the relationship linker, information about which members belong to which entities must be available. Use the Link Relationships job to run the relationship linker.
For the batch relationship linker, this information must be available on disk in the form of BXM files. To produce these BXM files, use the following procedure:
- Determine which relationships and which member and entity types need to be processed. Each relationship references one or two entity types. Each entity type is associated with a single member type.
- Create BXM files for each member type. You can use any of the MDM Workbench jobs that prepare derived data on disk: Prepare Binary Files (mpxprep), Derive Data from operational server (mpxredvd), Derive Data from UNLs (mpxfsdvd), or Derive Data and Create UNLs (mpxdata). If you are processing more than one member type, be sure to keep the output directories separate. Also make sure that the Generate Query BXM option is enabled.
- Generate the query UNL files by using the Derive Data and Create UNLs (mpxdata) job. Make sure that you set the Generate query UNL option. Load the query UNL files by using the Load UNLs to DB (madunlload) job.
- Create entity linkage information. For each entity type, run the Compare Members in Bulk (mpxcomp) job, followed by the Link Entities (mpxlink) job. Be sure to use the appropriate BXM directory for the entity type (the BXM directory of the associated member type you used in step 2.
- Now you are ready to run the batch relationship linker. The input
parameters and options and their meanings are:
- Inputs and Outputs tab
- Input directory
- These are the BXM directories you created in step 2, one for each member type. Separate multiple directories with a comma.
- Output directory
- This is the directory that receives the output UNL files: mpi_rellink.unl and mpi_relxtsk.unl.
- Relationship type
- To process all relationships, check Select All. To process a subset of relationships, clear Select All and select the relationships you want to process using the selection box.
You can also run the batch relationship linker by using the run_rellinker target of the madconfig utility. If you run the batch relationship linker in this way, you are asked for an additional parameter:
- Performance Tuning tab
- Buffer Size
- This controls the amount of memory used during relationship processing. The default value (0) can be used for most cases. The value indicates the number of rows stored in memory. Row sizes vary, but are approximately 256 bytes.
- Max Merge Degree
- This limits the number of open files used during relationship processing. The default value (0) indicates that the number of open files is not limited. Using a value other than the default causes relationship linking to be slower. On most operating systems, you do not need to modify this value. Modify this value only if batch relationship linking fails with a too many open files error.
- Maximum JVM heap size
- This controls the size of the heap used in the Java™ Virtual Machine (JVM). (See the description of the -Xmx option of the Java program.)
- Inputs and Outputs tab