BGP Analytics with OpenConfig Telemetry and gRPC

Peter Sievers

Playlists: 'denog11' videos starting here / audio / related events

Extracting BGP Telemetry Information from Network devices using vendor-neutral Openconfig data model and gRPC. Importing BGP metrics like received and advertised routes, total prefixes, session flap and more into TICK Stack and using mathematical models to analyse the data to identify anomalies.

This talk will discuss the benefits of streaming telemetry in conjunction with vendor independent data models with Openconfig and structured data for analysis versus SNMP and CLI screen scraping.
The further processing of the imported metrics is shown using the TICK Stack and Grafana as a dashboard,
for an optical visualization of the metrics, the primary focus is on BGP metrics.
Having all this metrics collected one can define statistical models to analyse or to correlate the data.
For example to use algorithm like 3-sigma rule or k-means to identify anomalies. Kapacitor together
with user-defined-function can do the job.
To Sum up, collecting different metrics from network devices can help the operator to have some kind of "network fingerprint", with this fingerprint the operator is able to identify anomalies. To easy analyse data you need it in a vendor-neutral structured format and therefore Openconfig data model is a good match.