4.5 KPI Management Agents

The KPI Cluster In and KPI Cluster Out agents are real-time processing agents that deliver KPIs based on the configured service model and the incoming KDR UDRs.

The KPI Cluster In agent sends incoming KDR UDRs to a Spark application via a Kafka cluster. The actual KPI calculations are then performed in the Spark Application and the result is passed to a KPI Cluster Out agent, which decodes the bytearray payload from Kafka and delivers KPIAggregatedOutput UDRs. 


KPI Management workflows


When the service model is configured to trigger immediate alarms you must configure two separate Kafka Collector agents, one for the kpi-output topic, and another for the kpi-alarm topic.

Configuration

KPI Cluster In configuration


KPI Cluster Out configuration


SettingDescription
Kafka ProfileClick  Browse to select a predefined Kafka profile. The profile contains the service model configuration.
KPI ProfileClick  Browse to select a predefined KPI profile. The profile contains the service model configuration.

Input/Output Data

The Input/Output data is the type of data an agent expects and delivers.

The KPI Cluster In agent expects KDR UDRs and delivers KafkaUDRs.

The KPI Cluster Out agent expects KafkaUDR types and delivers KPIAggregatedOutput UDRs.

MIM

For information about the MIM and a list of the general MIM parameters, see 1.2 Administration and Management in the Desktop User's Guide.

Publishes 

MIM ParameterDescription
Model Config

This MIM parameter contains the name of the KPI Profile in the agent configuration.

Model Config is of the string type and is defined as a global MIM context type.

Model Version

This MIM parameter contains the version of the KPI Profile in the agent configuration.

Model Version is of the int type and is defined as a global MIM context type.

Accesses

This agent does not access any MIM parameters.

Agent Message Events

There are no message events for these agents. 

Logs pertaining to KPI Management processing are available in Spark. For further information, see 4.7.6 Logging.

Debug Events

There are no debug events for these agents.