Configure service policies without response time goals

Configure service policies without response time goals when multiple service policies are configured and the response time goals cannot be accurately determined for any of the service policies, or when there might be occasional long-running or stuck requests.

To aid in the ease of use of the product, performance management without response time goals is supported. Providing correct response time goals requires load testing to be performed to determine a reasonable value. This feature eliminates the need for performance testing. Instead, the product simply uses the importance of a service policy to manage performance.

To determine when a new dynamic cluster instance should be started when there are no response time goals, a utility value is calculated based upon the fraction of time that requests wait. The utility value is a value between -1 and +1 based on the importance of the service class. Requests with zero waiting time get a +1 utility value, and requests with almost all time waiting get a -1 utility value.

A utility value of zero is obtained if the fraction of time waiting equals the relative importance of the request service class, where relative importance is zero for very high importance and one for very low importance. Hence, the utility value is a function of the performance level achieved, which is measured as fraction of time waiting, and the importance of the request, which is measured as relative importance. For a given importance level, a better performance level means higher utility values. When the utility value is less than zero, the application placement controller attempts to start a new instance if capacity to do so can be found.

  1. In the console, select System administration > Cell > Custom properties > New.

  2. Specify the name of the custom property as disableResponseTimeGoals.

  3. Specify the value of the property as true.