Metrics

Metrics in TrueSight Capacity Optimization are definitions of time-varying information about entities, either domains, systems, or business drivers. (See Entity Types for the various types of entities in TrueSight Capacity Optimization.) The actual information, in the form of metric instances, is loaded into TrueSight Capacity Optimization by connectors.

For more information, see the following sections:

Types of metrics

Metrics are of the following types:

  • Time series metrics: An instance of this sub-type contains a series of numeric values over regular time intervals, e.g., one value every five minutes, or one value every hour, etc. Each value represents something measured over that time interval (five minutes, one hour, etc.). Each interval of time can have at most one value in the series.
  • Configuration series metrics: An instance of this sub-type contains a series of values that have variable-length validity periods. Each value represents something that holds true during a time period t1 to t2. The validity periods in the series are non-overlapping, so that at any given instant in time, there can be at most one value valid. This kind of representation is suitable for configuration information, for example, hardware description or total amount of memory in a system, which does not change very often. Each value is represented as a string.
  • Events: An instance of this sub-type denotes a planned change or incident occurring about an entity at a specific time. To understand how these are used, see Events.

Datasets and metrics

TrueSight Capacity Optimization contains definitions of thousands of metrics, sometimes called resource counters, and in addition, a connector may define its own custom metrics. Every metric known to a TrueSight Capacity Optimization instance must have a unique string name called the "object" or "resource" identifier, conventionally in an all-capitals and underscores format, for example, CPU_UTIL, CPU_MODEL, and BYDISK_UTIL. As explained in Viewing datasets and metrics by dataset and ETL module, metrics in TrueSight Capacity Optimization are arranged in homogeneous groups called Datasets.

For example:

  • The dataset named SYSCNF contains over a hundred different Configuration series metric definitions, including CPU_MODEL and CPU_NUM.
  • The dataset named SYSGLB contains over a hundred and fifty Time series metric definitions, including CPU_UTIL.
  • The dataset named EVDAT contains a single definition for all Events.

One particular dataset, named OBJREL, contains a single definition for all relationships. For more information about relationships, see Entity relationships.

Custom metrics created by users or connectors are distinguished by a _C suffix in their name.

Metric instances and sub-object metrics

Datasets contain only the definitions of metrics. Instances of metrics are created by a connector based on the connector's data source.

Each instance of a metric is associated with only one entity. For example, a connector may create an instance of the time series metric CPU_UTIL and associate it with a particular system named Host1. Usually, there can be only one instance of a metric associated with an entity, as in the example of CPU_UTIL.

But some metrics are sub-object metrics. For such a metric, multiple metric instances may be associated with a single entity, one per sub-object of an entity. By convention, such metrics have names starting with the prefix BY. For example, the time series metric BYDISK_UTIL can have multiple instances for a system, one per disk. The connector that creates the metric instances assigns strings to each disk of each system, based on the data source. For example, if a system Host1 has two disks named diskA and diskB, then the connector may choose to create two instances of the time series metric BYDISK_UTIL, one instance for a sub-object of Host1 called diskA, and another instance for a sub-object of Host1 called diskB.

When looking at the metric instances of an entity in the TrueSight Capacity Optimization console workspace, you will see the metric and the sub-object of each metric instance identified in columns named Resource and Subresource, respectively.

LOCATION field in metrics used for business drivers

There is another way to associate multiple metric instances with a single entity: every metric also has a LOCATION field, by default filled with the string UNKNOWN. But when creating an instance, a connector may fill this field with arbitrary strings like Boston or New York. Then each such LOCATION value added, creates a separate instance of the metric associated with the same entity.

The intent of this field is to enable the connector to associate separate instances with geographic locations. Such a scheme might make sense for business drivers, for example, BYPAGE_RESPONSE_TIME could denote response times broken down by web page as well as by the location from where the page request was made. In our example, these two instances would be created for a single business driver entity:

  • BYPAGE_RESPONSE_TIME for sub-object ftp.bmc.com, for location Boston.
  • BYPAGE_RESPONSE_TIME for sub-object ftp.bmc.com, for location New York.

This use of the LOCATION field in metrics is rare, but it is available for when there are an unpredictable number of geographic locations that need to be tracked separately for a single business driver. You can achieve the same effect by creating separate business drivers, one for each location.

When looking at the metric instances of an entity in the TrueSight Capacity Optimization console workspace, you will see the LOCATION value identified in a column named Location.

Assigning a LOCATION field creates a new instance of the metric on the same entity! Do not use the LOCATION field if you simply need to associate a location with a system. Use a configuration metric for that purpose.

Metric instances associated with domains

Domains, i.e., entities of category APP, may have only configuration series metrics associated with them. Otherwise, domains behave just like the other two categories of metrics, namely systems (SYS) and business drivers (WKLD), for the purposes of metric instance creation.

The dataset named APPCNF contains definitions for a handful of built-in configuration series metrics, and users or connectors can define additional custom configuration series metrics.

Structure of a metric instance

A metric instance is identified by the following tuple:

Element

Meaning

ENTITY

Unique identifier for an entity, for example, a System ID for a system

ENTCATNM

Whether the entity is a domain ("APP"), a system ("SYS"), or a business driver ("WKLD")

OBJECT

Metric unique name, for example, BYPAGE_RESPONSE_TIME

SUBOBJECT

Sub-object name, for example, www.foo.com/index.html

LOCATION

Location field, for example, "Boston"

Each of these elements can be specified by a connector. The most typical case is for a system time series or configuration series metric, where the SUBOBJECT is GLOBAL and LOCATION is UKNOWN.

Metric values and summarization

Time series metric values are summarized by the TrueSight Capacity Optimization data warehouse according to the meanings of the metrics. The meanings are denoted by metric type codes, as follows:

Metric type ID

Metric type

Valtype

Meaning

Summarization

Examples

1

COUNT

A count of events, absolute number

A numeric value that indicates the count of events. For example, errors, transactions, and calls. 

SUM VALUE

Pages downloaded, errors

2

ELAPSED

Elapsed time

A numeric value that indicates the time required for a specific action or task to complete.

AVG VALUE

Response time

3

RATE

A frequency, in events/sec

A numeric value that indicates events or operations that occur per second.

AVG VALUE

Disk IOPS

4

PERCENT

Percentage counter

Numeric metric expressed in percentage.

AVG VALUE

Burst percentage, Memory
utilization percentage 

5

CONF

Configuration data

A numeric or textual value which indicates a configuration detail that does not frequently change such as Asset ID, Reference date, or max bandwidth.

(none)

Number of CPUs

6

POSACCUM

Positive accumulation counter

Same as peak counter.

MAX VALUE

Disk space used

7

NEGACCUM

Negative accumulation counter

A counter for which minimum value is of interest. For example, free disk space, free storage space volumes, or CPU.

MIN VALUE

Disk space free

8

GENERIC

Generic counter, absolute value

A numeric value for counter metrics. For example, power consumption or CPU service units.

AVG VALUE

CPU queue length

9

WEIGHTED_GENERIC

Generic counter, absolute value, weightedA numeric value for weighted counter metrics, where the weight is set in the ETL.WAVG VALUEWeb response time
10

PEAK_PERCENTAGE

Peak percentage counterPeak value for percentage metrics.MAX VALUEMemory Utilization Peak
Percentage, CPU Utilization
Peak Percentage
11

PEAK_RATE

Peak counterA numeric value that indicates positive peak values for events or operations that occur per second.

MAX VALUE

Peak packets received per second
12DELTADifference between subsequent samplesA positive numeric value that indicates the difference in value between two samples of a metric.SUM VALUE-
13WEIGHTED_PERCENTAGEPercentagecounter,weightedNumeric value expressed in percentage, that is weighted by a value imported by the ETL.WAVG VALUEEvents of a specific set,
weighed by total events expressed in percentage.
14

PEAK

Peak counterPeak value for counter metrics.

MAX VALUE

Peak download time
for a web page
15HIGHMARKHighmark counter, peak on granular samples A numeric value that indicates peak values on the granular samples.MAX VALUE

The TrueSight Capacity Optimization data warehouse uses the indicated summarization methods to calculate hourly, daily, and monthly values from the raw numbers loaded by connectors. No summarization is done for CONF metrics.

Where to go from here

Entity relationships

Was this page helpful? Yes No Submitting... Thank you

Comments