Monday, December 9, 2013

Performance Monitoring and Baselines


When they are properly set, alert thresholds provide a valuable service—an alert—by indicating a performance metric that is at an unexpected value. Unfortunately, in many cases the expected value varies with the workload type, system load, time of day, or day of the week. Baselines associated with certain workload types or days of the week capture the metric values of that period. The baseline can then be used to set the threshold values when similar conditions exist.

The statistics for baselines are computed to place a minimal load on the system; statistics for static baselines are manually computed. You can schedule statistics computation on the AWR Baselines page. Statistics for the system moving window are automatically computed according to the BSLN_MAINTAIN_STATS_SCHED schedule. By default, this schedule starts the job every week at noon on Saturday.

Performance Monitoring and Baselines
Metric statistics computed over a baseline enable you to set thresholds that compare the baseline statistics to the current activity. There are three methods of comparison:

  • significance level
  • percentage of maximum and 
  • fixed values
Thresholds based on significance level use statistical relevance to determine which current values are unusual. In simple terms, if the significance level is set to .99 for a critical threshold, the threshold is set where 1% of the baseline values fall outside this value and any current values that exceed this value trigger an alert. A higher significance level of .999 or .9999 causes fewer alerts to be triggered.

Thresholds based on percentage of maximum are calculated based on the maximum value captured by the baseline.

Threshold values based on fixed values are set by the DBA. No baseline is required.




No comments:

Post a Comment