Optimization of CPU Overhead for z/OS Java

moritzwitti — Thu, 11 Jun 2026 14:54:59 GMT

Summary

This article applies to Dynatrace OneAgent in mainframe (z/OS) environments and addresses increased zIIP usage related to SQL and database query processing. It explains the cause and describes configuration options introduced to reduce CPU overhead and improve SQL processing efficiency.

Problem

Customers may observe an increase in zIIP usage, particularly in environments with SQL‑ or database‑intensive workloads. The increase is typically correlated with JDBC-based SQL execution and can raise concerns about higher CPU consumption and performance impact.
If a customer experiences increase in their zIIP usage especially in the scope of SQL/DB queries, we introduced two features that should result in performance improvements.

Troubleshooting steps

Identify whether the zIIP usage increase correlates with SQL or database activity.
Verify the Dynatrace OneAgent version running in the affected environment.
Check whether SQL normalization and PreparedStatements reuse are relevant for the observed workload.

Resolution

Dynatrace introduced SQL processing optimizations to reduce CPU overhead and improve performance in SQL-heavy workloads.These optimizations include:

Improved SQL normalization efficiency through optimized SQL tokenization
A PreparedStatements cache to reduce repeated parsing and normalization

To enable these optimizations, configure the following settings on the affected OneAgent:

jdbc-sql-optimize-tokenization = true
jdbc-sql-cache-prepared-statements = true

Both settings are available starting with OneAgent release 327 and can be set either via

extending the dtconfig.json with "jdbc-sql-optimize-tokenization": true or
adapting the JVM arguments to include the two arguments such as -javaagent:/PATH_TO/dynatrace-oneagent-zos-java.jar=jdbc-sql-optimize-tokenization=true,jdbc-sql-cache-prepared-statements=true

Internal testing shows the most significant improvements in environments where PreparedStatements are effectively reused.

What's next

Monitor zIIP and CPU usage to validate the impact
Review application behavior to ensure PreparedStatements are reused where possible
Continue performance analysis using Dynatrace to confirm reduced SQL processing overhead