Will Post Enhancement Performance (PEP) work in a Change History or Change Data (CDC) environment? (4296856)

SharePlex offers an array of configuration choices to suit one’s replication needs. At the same time numerous features are added in SharePlex from time to time. Some of these features may not work in all replication topology. This article addresses the specific question of support for PEP, Post Enhanced Performance, in a CDC environment.

The PEP, Post Enhanced Performance, feature will work on a CDC environment as a CDC environment as a CDC environment merely involves INSERTs to the target database. Enabling PEP will only benefit if the  transactions  on source are small. Even if the logging of COMMITs is suppressed by enabling the parameter SP_OPO_TRACK_COMMITS on target, the PEP feature will still work. The reason being, the SP_OPO_TRACK_COMMITS only suppresses the logging of COMMITs but the Post still has to club the COMMITs if the PEP is in place so the two do not interfere. 

Additional Information:

The PEP has two components, concurrency and commit reduction. The concurrency is not relevant in a CDC target as there are no constraints so checking for dependency is not applicable. Only the commit reduction feature will be applicable when talking about PEP in a CDC environment. The knowledgebase article 229046 delves mainly on commit reduction feature of PEP but also has a passing reference to concurrency and the following is an excerpt from that article:

SharePlex Release Notes have the description of the feature as below:

Post performance enhancements

The new Post Enhanced Performance (PEP) feature enhances Post to provide better performance when applying data to Oracle targets. Options were added to reduce the number of commits and to improve throughput by increasing the level of concurrency. Use the PEP feature to improve the speed of Post when it is processing mostly small transactions, such as those most commonly found in OLTP.

PEP has two levels:

Increase concurrency

Reduce the number of commits

Increase the level of concurrency

You can configure Post to apply transactions in parallel to increase overall throughput. The smaller the transaction, the bigger the performance gain. To enable the concurrency feature, set the new SP_OPO_DEPENDENCY_CHECK parameter to 1.

Reduce the number of commits

You can configure Post to combine smaller transactions into larger ones. This reduces the number of commits and acknowledgements that must be processed. The smaller the transaction, the bigger the performance gain.

Commit reduction is enabled by default when you enable the concurrency feature, and it cannot be enabled without enabling concurrency. The new SP_OPO_COMMIT_REDUCE_MSGS parameter controls the size of the combined transaction. It sets the minimum number messages (operations) in the transaction. Post skips the commits of small transactions whose boundaries are within the specified range, and instead applies all of the operations in the combined transaction. The default combined transaction size is 100 messages. To disable commit reduction, set this parameter to a value of 1.

The SP_OPO_COMMIT_REDUCE_MSGS setting is not an absolute threshold. SharePlex will not break up a transaction across different combined transactions. Therefore, Post may need to exceed that threshold in order to include all of the operations and the commit of the last transaction in the group.

With both concurrency and commit reduction enabled, testing has shown that performance can be increased by as much as two or three times over conventional SharePlex posting speeds.

Note: The PEP feature may reduce or eliminate the need to run multiple Post processes, but you can still benefit from the use of multiple Post processes because they eliminate a single point of failure. If a Post process fails, the other Post processes will continue, resulting in less recovery time after the problem is resolved. This feature can be used in a multi-Post configuration, so long as the rules for using multiple processes are followed (such as including all tables with referential integrity in the same process stream).