Performance Investigator (PI) displays historical data based on a predefined time pyramid, which controls data resolution and how long it is kept.
PI keeps detailed data for a limited time. As time passes, data is rolled up into coarser intervals. The current data retention periods are fixed and cannot be changed.
| Time Resolution | Retention Period |
|---|---|
| 1 minute | 6 hours |
| 15 minutes | 3 days |
| 1 hour | 2 weeks |
| 6 hours | 30 days |
| 1 day | 90 days |
| 1 week | 2 years |
For example:
Data older than 6 hours will only be available at 15-minute resolution or higher.
After 3 days, 1-minute and 15-minute data is no longer available.
PI onlys sample top 500 records each second, and submit top 300 unique combinations (batch,user,database,program etc) analysed from the sample records each minute. So if there're more than 300 unique queries each minute, then some loss of data is inevitable, The design is meant to gather the top activity most affecting the workload as these are the things that can have the most impact if tuned. It is not aimed to include all information about all workload. When hundreds of queries are run there are more than 300 batches running each minute, some data will likely be missed.
During data rollup, PI selects top records based on the data type, not all records. As a result:
Some queries or activity may not appear in higher-level summaries.
Active time may look reduced in rolled-up views.
Note:
On the instance level, Foglight keeps total active time consistent and does not reduce raw data.
