The Case for Continuous Performance Monitoring

In some installations, monitoring of performance activity is done on a demand basis. When a performance problem is reported, the performance analyst runs one or more commands in an attempt to determine why the problem occurred. In some cases, explicit recreation of the problem is needed in order to collect analysis data. The result is that users experience every performance problem twice.

It is usually more effective to monitor performance continuously, preferably with automatic collection of additional data if performance deteriorates. The costs of continuous monitoring are outweighed by the advantages, such as:

• Monitoring can sometimes detect incipient problems before they have an adverse effect.
• Monitoring can detect problems that happen to users who are reluctant to complain and problems that are not quite severe enough to complain about--but are affecting productivity and morale.
• Monitoring can collect data when a problem occurs for the first time.

Successful monitoring involves the following main activities:

• Periodically obtaining performance-related information from the operating system
• Storing the information for future use in problem diagnosis
• Displaying the information for the benefit of the system administrator
• Detecting situations that require additional data collection or responding to directions from the system administrator to collect such data, or both
• Collecting and storing the necessary detail data

The following sections discuss several approaches to continuous monitoring. These approaches are not mutually exclusive, but use of more than one may involve some redundancy.

• "Performance Monitoring Using iostat, netstat, vmstat"
• "The Performance Diagnostic Tool"