Guidelines for Specifying Data Reporting Intervals Junos Telemetry Interface
The Junos telemetry interface enables you to provision sensors to collect and export data for various system resources without involving polling. A request to send data is sent once by a management station to stream periodic updates.
You can configure telemetry sensors to report data at a specified
interval either through the command-line interface (CLI) or through
the OpenConfig for Junos telemetrySubscribe remote procedure
call (RPC). To configure using the CLI, include the reporting-rate seconds statement at the [edit services analytics export-profile profile-name] hierarchy level. For the telemetrySubscribe RPC, specify the sampling interval parameter,
in milliseconds. In both cases, the interval specifies the amount
of time between each subsequent export of data.
How to Determine the Reporting Interval for a System Resource
To determine the appropriate reporting interval for a specific system resource, follow these guidelines:
Identify the required export interval for a given object, such as an interface.
Identify the maximum number of objects reported by the sensor, such as the number of physical interfaces configured on a line card.
Identify the minimum number of objects reported on each interval for a given sensor.
Use the following formula to determine the best reporting interval:
Reporting interval = Required Export Interval Per Object * Minimum Number of objects reported on each Interval / Maximum Number of Objects.
Consider this example. There is a business requirement to report interface statistics every 30 seconds. At every interval, 10 interface records are reported, and the total number of interfaces is 96 for each line card. Using the reporting-interval formula, the reporting interval should be 3.125 seconds. Currently, the reporting interval can be configured only as a multiple of 2, in seconds. Therefore, for this example, configure the reporting interval as 2 seconds in the CLI or 2000 milliseconds in the OpenConfig RPC.
The same metric might be reported more than once over a 30-second interval. For the purposes of effective visualization and data manipulation, it is quite common to aggregate data over fixed time spans.