Inline Flow Performance Monitoring (FPM) complements BFD in measuring peer paths between two SSR devices. To provide this additional level of detail, the initiating router adds a small amount of metadata to a sample number of packets. This metadata indicates that metrics are collected per path, traffic class, and protocol (TCP or UDP). The receiving router reads the metadata and returns the data to the intiating router. The configurable fields include:
name: The name of the performance monitoring profile.
marking-count: The number of packets to mark within a given interval.
interval-duration: The duration of a packet marking interval in milliseconds.
monitor-only: Collect statistics without influencing packet processing features.
resource-group: Associate this performance monitoring profile with a top-level resource-group.
Additional information is available at
The data returned is organized by traffic class (high/medium/low/best-effort) and protocol (TDP, UDP), giving more granular data on the performance of a link - not just through synthetic traffic (BFD), but by piggybacking onto actual user traffic.
The SSR has four traffic classes, in order of priority: high, medium, low, best-effort. The traffic-class allows you to configure different treatments for different classes of traffic for the same service.
Working in concert with the traffic engineering subsystem, this lets you determine which paths the traffic referenced by a service class will use, and therefore how much guaranteed bandwidth is reserved for each class.
- From the Authority level, scroll down to Performance Monitoring Profiles, and click ADD.
- Enter a name for the profile.
- Configure monitoring values, or accept the defaults to gather a baseline.
The default setting for Monitor Only is false. This generates metrics, but does not influence load balancing decisions or traffic flow. To use the Flow performance metrics for load balancing decisions, select true.
Optionally, assign the profile to a Resource Group.
Validate and Commit the configuration changes.
Additionally, ensure that the profile is applied to your peer paths. This is generally configured through router > node > device-interface > network-interface > neighborhood. Apply the profile to both sides of the SVR peer path.
Flow performance monitoring operates on top of existing traffic, and does not intiate its own traffic. Because of the interval timing and sample time, no significant load is added to the path, and no impact to load balancing is generated.
By default, the metrics generated by FPM are not contributed to the BFD metrics to influence load balancing. By disabling monitor mode (set to true) the FPM metrics are reported to the load balancer.
When FPM is active (not configured as monitor only), both BFD and FPM provide metrics for calculating SLA and making load balancing decisions. Because FPM metrics provide greater detail than BFD metrics, these metrics are used first by the load balancer. When FPM metrics are stale (due to lack of traffic) the use of BFD metrics resumes.
show stats performance-monitoring to view the results. The following metrics are provided:
peer-path jitter: Jitter in milliseconds for the SSR peer path.
peer-path latency: Latency in milliseconds for the SSR peer path.
peer-path loss: Packet loss percentange for the SSR peer path.
peer-path mos: MOS value calculated for the SSR peer path (hundreths of a decimal).
peer-path updated: Time of last reported FPM activity on this peer path.
Verify that the peer router has FPM enabled on its peer path. If not, the peer router will not collect and respond to FPM data analytics queries.
FPM operates on packets below MTU so calculations are not impacted by fragmentation stacks. Certain test tools that by default generate jumbo traffic may not satisfy this condition. As a result, FPM will not sample packets that will not generate new analytics.
Because FPM attaches to existing traffic and does not generate its own, a very short interval duration can lead to inaccurate loss values. This is due to FPM not being able to sample a sufficient number of packets within the given interval. Increasing the interval duration allows more packets to be sampled, providing a more accurate representation of loss on the path.