TB119 Performance Regression May Occur When Masking String Data

Impact

Continuous Compliance Engines may see significant performance reductions in database masking jobs that operate on certain String data types. Worst-case performance of masking jobs is 25% of normal, i.e., masking jobs may take up to four times as long.

Correctness of masking results is not impacted.

This issue does not affect the performance of file or mainframe dataset masking.

Contributing Factors

This article applies to the following versions of Delphix engines:

Only Continuous Compliance Engines are affected by the issue. 

The issue may affect a large spectrum of jobs involving masking of string data such as varchar types. Character String data encoded in CHAR, VARCHAR, or TEXT (non-Unicode) is affected. Strings encoded in Unicode are not impacted, nor are non-String types such as DATE and INT.

The issue is known to affect all Advanced and Basic connectors. 

Oracle and MS SQL Server connectors are known to be affected by the issue, and it's presumed that a broad collection of other connector types could be similarly affected. 

Performance degrades as more columns are involved in masking affected String types. For example, masking ten such columns may result in twice the performance degradation as masking four columns.  

Symptoms

Jobs may take significantly longer than they took before upgrading to an affected release.

Relief/Workaround

Defer upgrades of Continuous Compliance engines to any of the affected releases. 

For systems already running an affected release, upgrading to 22.0.0.1 or later will correct the issue.

Related Documents

N/A