Oracle database has two different categories of performance data: cumulative statistics and sampled data.
Good examples of cumulative statistics are those available in DBA_HIST_SYSSTAT view. Cumulative statistics are
information summed up at database instance level. Their value keeps growing as database activities go on and will
be reset when database instance restarts. When cumulative statistics are examined, the delta value between two points
of time is calculated and the amount of change is measured. Metrics are derived from cumulative statistics by calculating
the rate of their changes against a time unit, or other units like transaction, database call etc.
Good examples of sampled data are those available in DBA_HIST_ACTIVE_SESS_HISTORY view. Oracle uses ASH framework
to collect session level information by sampling active sessions on database. When sampled data is examined, their
percentage within each sampling set and their frequency across multiple sampling sets are calculated. The significance
threshold is therefore derived.
Because Oracle uses AWR as framework to save all performance data into DBA_HIST views, there is a tendency to
treat ASH as a component of AWR. However, Oracle provides AWR report and ASH report as two separate tools to analyze
performance data, which marks the clear differences between these two categories:
- AWR data is cumulative data at database instance level
- ASH data is sampled data at session level
From the perspective of Oracle database tuning, DBspeed considers the 'data' as performance data stored in both AWR and ASH.