Oracle AWR data storage and collection frequency optimization.

In the production Oracle environment, it is essential to change AWR retention and performance data snapshot collection frequency from 1-hour to 15 minutes. Our F500 major clients use 15 minutes with no impact on production performance. Having 15 minutes of data snapshots vs. 1 hour dramatically increases the ability to troubleshoot. Also, AWR data contains ASH information collected from memory. Switch to 15 minutes collection intervals dramatically improves the quality of ASH information.

Two weeks is the default period of retention. At least four months of AWR data is required in a production context compared to four months of monthly seasonal data. Compare one month from the current quarter to the same month from the previous quarter. As a result, storing new gigabytes of AWR data is not a budgetary issue. Also, we strongly suggest keeping 13 months of data to investigate annual seasonality.

Below command switches AWR to 4 months (172800 minutes) data retention and 15 minutes collection frequency.

exec DBMS_WORKLOAD_REPOSITORY.modify_snapshot_settings(retention => 172800,interval => 15);

Overall, keeping several years of AWR performance data for capacity planning and predictive analytics employing advanced machine learning technologies to discover and predict mid-and long-term performance trends and probable “Black-Swan” events. Events that have the potential to dramatically and catastrophically impair the production environment will hold.

For these purposes, Oracle introduced The Enterprise Manager AWR Warehouse. As a result, Enteros UpBeat fully supports database performance analysis across AWR Warehouses and individual production database systems.

Enteros’ expert team has deep expertise in predictive analytics on database performance data. Contact us at support@enteros.com, and we will gladly discuss how our experts can create machine and deep learning models to predict the performance of your critical database systems.