Compute Customized Metrics
Although Pepperdata provides hundreds of metrics—individual as well as metrics that are computed by performing arithmetic operations on one or more basic metrics—you might still want something more. You can use the charts’ advanced options to add charts to show custom computed metrics. Typical scenarios are: scaling individual metrics; performing arithmetic operations across series; and finding the ratio of two basic metrics.
Computed metrics are expressions written in RPN/postfix notation (see the Wikipedia article, Reverse Polish notation ). Your expressions must conform to the following rules:
Valid operators for computed metrics expressions are the four basic arithmetic operators, two inequality operators, and integral/derivative unary “operators” (the operation words themselves):
If an inequality relationship is satisfied between operands, the output is
1; otherwise the output is
Do not use parentheses. (They are unnecessary in RPN/postfix notation.)
Use the following sample expressions as models to build your own.
Average memory per task by user:
ucpu tasks /
Ratio of processes to tasks on a host:
n_ltp tasks /
Time spent in dynamic tasks as a ratio of overall tasks:
1 ndynall_base_size ndynall_current_size / - tasks * INTEGRAL tasks INTEGRAL /
The “subtract from 1” portion of the expression inverts the percentage so as to arrive at the correct ratio,
ndynall_base_size / ndynall_current_size, which in RPN/postfix notation is
ndynall_base_size ndynall_current_size /.
Number of files opened to write < 1 MB of data:
dfsbw INTEGRAL dfsfow INTEGRAL / 1000000 < dfsfow INTEGRAL *
Number of files opened to read < 1 MB of data:
dfsbrps INTEGRAL dfsfor INTEGRAL / 1000000 < dfsfor INTEGRAL *
Display a chart.
Click the Metrics filter bar.
In the Search section, select Add Metric Expression.
Enter the Computed Metric to Chart expression: metric names, operators, and/or constants.
To find a metric’s name:
If the metric is already charted, mouse over the chart’s “?” tooltip icon (); the metric name is in parentheses.
For the full list of metrics, refer to the Metric Definitions page on your dashboard (on the navigation bar, click Help > Metric Definitions). A metric’s name is given in parentheses after the friendly name.
In the following example, the friendly name is “user CPU percentage”; the metric name is
c_ucpupct; and the description is “Percentage of user CPU time for the monitored component”.
user CPU percentage (c_ucpupct) Percentage of user CPU time for the monitored component.
Click Add, and click Apply.
The chart view refreshes to show your computed metric, as the last chart on the page.
To see the metric expression that you entered, mouse over the “?” tooltip icon () next to the chart’s title.