Poll intervals

The number of poll intervals affects the response times. A guideline is provided to help with choosing the appropriate interval.

You define the poll interval when you create profile elements. Use the following formula as a guide:

minimum poll interval = number of profile elements × average response time / MaxCCA

The number of profile elements configured for a monitor determines the total number of tests performed by the monitor. The average response time varies according to the monitoring environment, so when selecting the poll interval, choose a value appropriate to the application environment. In the worst case, if the application fails to respond then the average response time is the timeout value set for the monitored element.

Note: If you use history collection, use the same poll intervals as the history collection intervals.

Response times in LAN monitoring environments

In LAN monitoring environments, server tests run over high-speed networks. Therefore the response times of services are low, typically less than one second.

The poll interval is also affected by data logging functions that involve a higher number of disk access operations. If you use monitors as data sources for only IBM Tivoli Netcool/OMNIbus and IBM Tivoli Monitoring, and you do not use the datalogs as storage for simple reporting or archiving results, then monitor data logging is not necessary and you can run the monitors without data logging. This condition permits shorter polling intervals of up to one half of the value indicated by the polling interval formula.

Response times in remote monitoring environments

In remote monitoring environments, where service tests run over a WAN or an Internet connection, the response times are less predictable, and it is the network response time, instead of monitor performance, that limits the poll interval. Use polling intervals that allow enough time for responses to be received before the next test starts.