Capacity Agent Metric groups and metrics

This topic contains the following sections:

Related topic

Available metric groups

A metric is the measurement of a particular characteristic of performance or efficiency. A basic installation of TrueSight Capacity Optimization makes the following metric groups available:

SQL server application metrics (through the registry)
UNIX and Linux metrics
Windows NT, Windows XP, Windows 2000, and Windows 2003 Common metrics
Windows 2000-specific metrics
Oracle VM metrics

These sections serve as a reference that lists and describes the UNIX, Linux, and Windows system metrics that are collected and made available through TrueSight Capacity Optimization. The organization of these sections corresponds to the organization of the system metric groups in the Investigate interface.

Metric groups are displayed in the following contexts in the Gateway Server interface: when you create a filter, when you add a condition to an alert, and when you drill down.

For information about metrics available for all supported platforms, see:

Metric computation

The supported computations for a given metric depend on its data type. These computations are described in the following table, and they represent the post-processing operations performed by the Agent on the raw data.

Metric computations and supported data types

Computation	Description	Supported data types
% Time	Reflects the percent of the data collection interval spent in a given state. Examples: For the System CPU Time metric, % Time indicates the percent of the data collection interval the processor spent in a system state. For the Disk Active Time metric, % Time indicates the percent of the data collection interval the disk was busy.	Time
Delta	Reflects the metric change during the data collection interval	Integer Count Float
% Change	Reflects the percentage of change in the raw metric value during the data collection interval	Integer Count Float
Rate of change	Reflects how fast the raw value of the metric is changing over a specified period of time	Integer Count Float
Raw	Indicates that no computation is being performed on measured data	All

Note

You specify the data collection interval in the Alert dialog box. The agent uses this value to schedule data collection. For computations, however, the agent uses the actual collection times to compensate for any scheduling irregularities.

Metric computation equations

The following symbols are used for the metric computation equations, as shown in the following table:

PV - Previous value of the metric
CV - Current value of the metric
PT - Collection time of previous data point
CT - Collection time of current data point

Metric computation equations

Computation	Equation
% Time	`100 * (CV-PV)/(CT-PT)`
Delta	`CV-PV`
% Change	`100 * (CV-PV)/PV`
Rate	`(CV-PV)/(CT-PT)`

Metric data types

Metrics are classified into data types based on how the data is collected by each data collector. The data types determine how TrueSight Capacity Optimization treats the associated metric data, as well as how the Capacity Agent computes the data.

The following table lists the data types in alphabetical order and provides a brief description and example.

Metric data types

Data Type	Description
Character	Data type for text strings. Character strings can be compared for alerting and filtering purposes. Cannot be computed.
Count	Metrics cumulative since system boot time or process invocation
Float	Usually reflects data useful in its collected (raw) form. Can be further computed to detect trends. Example: File System Utilization is collected as a floating-point percentage of the total file system capacity. Alert conditions can be specified based on the raw percentage of capacity currently being utilized. You can also monitor growth rate of the file system, where your alert indicates whether the file system will soon be full. Example: An alert can specify that the system will be full in 10 seconds.
ID	Identifier stored as a number, such as process ID. Cannot be computed.
Integer	32-bit whole number whose size is not dependent on machine word size. Values that are usually useful without computation. Can be computed to determine trends. Example: Free Memory is itself a useful metric to monitor without computation. Add computation Rate of Change, and it can also predict when your system will run out of memory.
Time	Cumulative count of time in seconds. Similar to the Count data type, but used differently. Used to measure the amount of time a processor spends in a certain state, a disk spends being active, or a process spends in a system or user state. Useful only as a percentage of the data collection interval. Example: System CPU Time is collected as a total number of seconds since system boot that the processor has spent in the system state. Computation % Time (default) reports the percentage of the last data collection interval the processor spent in the system state.
UTime	UNIX time format. Distinct from time value format in that it is primarily used as a time stamp. (UTime has been represented in seconds since January 1, 1970.) Always denotes specific date and time rather than an interval length. Example: Process Start Time, which displays the time and date a process was started

Metric data units

Metrics are collected in a raw data format of base units and require further processing to produce meaningful results.

The base units available are determined by the authorized computations for a particular data source. TrueSight Capacity Optimization provides default units for each metric.

For each data type, the following table shows the base units, default units, and available units, and the relationship between data type's units and computations.

Relationship of data types to unit computations

Data type	Base units	Default units	Available units	Default computation	Available computations
Count	units	units/Second	units units/Sec units/Interval	Rate of change	Raw Rate of change Delta
Float	units	units	units units/Sec units/Interval % Change	Raw	Raw Rate of change Delta % Change
Integer	units	units	units units/Sec units/Interval % Change	Raw	Raw Rate of change Delta % Change
Time	Seconds	%	% Seconds Seconds/Interval	% Time	% Time
UTime	Seconds	Seconds	Seconds	Raw	Raw