SDN Controller Programming Guide

Figure 19 Metrics Architecture

Essentially a component or application must contact the MetricService to create a new

TimeStampedMetric on their behalf; they will be returned a reference to the resulting (new)

TimeStampedMetric object. The developer can then manipulate the returned TimeStampedMetric

object as appropriate for their own needs, updating its value at their own cadence, on a regular

or irregular basis, to reflect changes in whatever is being measured.

Behind the scenes, the MetricService API is backed by a MetricManagerComponent OSGi

component. This component delegates almost all of its work to a MetricManager singleton, which

(conceptually) contains a centralized Collection of the TimeStampedMetric references doled out at

the request of other components and applications. This Collection of TimeStampedMetric

references allows the metering framework to process the TimeStampedMetrics In masse,

irrespective of which application or component requested them, in a fashion that is completely

decoupled from the requesting application's or component's use of the TimeStampedMetrics.

The most essential processing done by the metering framework is to periodically persist

TimeStampedMetric values to disk, and to expose "live" TimeStampedMetric values through JMX.

Other processing is also done, such as aging out old TimeStampedMetric values. Decoupled from

this ongoing persistence of TimeStampedMetric values that are still being used, values that have

already been persisted from TimeStampedMetrics over time may read via REST API and exported

for further analysis or processing outside the controller

TimeStampedMetric Types

There are seven types of TimeStampedMetric. They are listed below, with an example of how each

type might be used.

 TimeStampedCounter

 A cumulative measurement that is incremented or decremented when some event occurs.

 Example application: the number of OpenFlow devices discovered by the controller.

 TimeStampedGauge

 An instantaneous measure.

 Example application: the amount of disk space consumed by metric data.