Partitioning storage of execution logs

If HiRDB is used as the database to store a very large process instance execution log, you can use partitioning storage for the execution log to efficiently search and delete log data.

For Service Platform, you can use key range partitioning with boundary values specified using date data. You can also use matrix partitioning, in which partitioned data is further partitioned by FIX hash partitioning using process instance identifiers.

To use matrix partitioning, HiRDB Advanced High Availability is required.

For details about key range partitioning and FIX hash partitioning, see the explanation of the types of table row partitioning in the HiRDB Installation and Design Guide.

The following figure shows an example of matrix partitioning using a combination of key range partitioning with boundary values (date data) specified and process instance identifiers.

Figure 7-3 Example of matrix partitioning using a combination of key range partitioning with boundary values (date data) specified and process instance identifiers

[Figure]

Organization of this subsection: (1) Key range partitioning with boundary values specified using date data; (2) FIX hash partitioning using process instance identifiers; (3) Notes on dividing the execution log

(1) Key range partitioning with boundary values specified using date data

For Service Platform, the table column name (SplitKey) related to the management of process instance execution logs can be used as the key range partitioning key.

By dividing the execution log and storing it in RDAREAs in a row-partitioned table on the database server, you can manipulate log data in units of RDAREAs. Therefore, the larger the execution log, the more efficient deletions become.

For details about an example of deleting execution log data in units of RDAREAs, see 6.1.4 Deleting process instance execution logs in the Service Platform Setup and Operation Guide.

(2) FIX hash partitioning using process instance identifiers

For Service Platform, the table column name (ProcessID) related to the management of process instance execution logs can be used as the FIX hash partitioning key.

If the partitioning key is specified for the search condition, only the RDAREAs expected to have the target data will be searched. This increases search efficiency.

Note: If the partitioning key is not specified for the search condition, all RDAREAs will be searched. As a result, due to communication between servers, the searches might not be performed as efficiently as when partitioning is used.

The SQL statement that is issued during request processing differs depending on the business process contents. To check details about the SQL statement that is issued, use the SQL trace functionality. For details about the SQL trace functionality, see the HiRDB UAP Development Guide.

(3) Notes on dividing the execution log

Note the following when dividing the execution log for process instances:

You cannot migrate the execution log from an environment in which an HCSC server has already been set up.
To delete the execution log data in units of RDAREAs, you must stop reception on all HCSC servers.
Enhance the duplication check of correlation sets. To enhance the duplication check of correlation sets, specify ON for the db-tbl-split-corcheck-use property in the HCSC server setup definition file. If OFF is specified, requests might not be processed correctly. For details about the db-tbl-split-corcheck-use property, see HCSC Server Runtime definition file in the Service Platform Reference Guide.
If the same value is specified for correlation sets, all of the requests might result in errors. Specify a unique value for each correlation set.

7.3.1 Partitioning storage of execution logs

(1) Key range partitioning with boundary values specified using date data

(2) FIX hash partitioning using process instance identifiers

(3) Notes on dividing the execution log