Metrics, Alerts, and Baselines

A metric is a timed rate of change in a cumulative statistic. For example, physical reads per second. 

Threshold values can be set for various metrics, and when the value of the metric crosses the threshold value, an alert is generated.

Monitoring for performance requires certain information that goes beyond statistics. To determine whether a particular statistic is important, you need to know how much it has changed over a certain period of time. 

To be proactive, you need to be notified when certain conditions exist, for example when system response time approaches the agreed maximum. To diagnose performance issues, you need to know what has changed. Metrics, alerts, and baselines provide this information.

A metric is a timed rate of change in a cumulative statistic. For example, physical reads per second.

A metric is another type of statistic collected by Oracle Database. A metric is defined as the rate of change in some cumulative statistic. That rate can be measured against a variety of units, including time, transactions, or database calls. For example, the number database calls per second is a metric. Metric values are exposed in some V$ views, where the values are the average over a fairly small time interval, typically 60 seconds. A history of recent metric values is available through V$ views, and some data is also persisted by AWR snapshots.

Threshold values can be set for various metrics, and when the value of the metric crosses the threshold value, an alert is generated.

A baseline comprises stored sets of metrics and statistics. A single set is called a snapshot. A baseline is made up of two or more snapshots. 

Usually a baseline is captured during a period of normal or acceptable operation, but it can capture any period of interest. When performance is not as expected, another set of metrics can be captured and compared with the baseline. This method allows the data to clearly point to the performance issues.

The Oracle database server collects base statistics during normal operations. Base statistics are simple counts. For example, counting the number of physical reads in the system since startup is a base statistic.

Metrics are secondary statistics derived from base statistics. Most metrics track the rates of change of activities in the Oracle server. For example, the average physical reads in the system in the last 60 minutes is a metric. Metrics are used by internal components (clients) for system health monitoring, problem detection, and self-tuning. The Manageability Monitor process (MMON) periodically updates metric data from the corresponding base statistics.

Oracle Database 11g supports metrics for system, sessions, files, and wait event statistics. Each metric is uniquely identified by a metric number and is associated with a metric name.

Benefits of Metrics

• The main benefit of keeping metrics is that when a component needs to compute the rate of change of some activity, the data is readily available.
• In earlier releases, you had to capture statistics before and after running a workload to compute the changed rate for a particular base statistic. With metrics, all you need to do is run your workload and select the corresponding metrics.
• Metrics provide the performance information required for the Automatic Memory Management and Automatic Database Diagnostic Monitor.