8-9 September 2020
Europe/Prague timezone

BGP Telemetry

8 Sep 2020, 11:20


CSNOG 2020


Telemetry is a universal marketing buzzword covering all aspects of interconnections – all but BGP, at least for now. This is not ideal and needs to be rectified. :-)

Two interrelated topics are proposed – BGP Telemetry and Peering Telemetry, the first one being focused more on the technical aspects of BGP telemetry mechanisms based on BMP, and the second one is more on a higher level overview and value proposition of IX telemetry, with the focus on interworking of BGP and data plane telemetry mechanisms. The actual topic could be adjusted between the two options to better suit the intended audience.

Title: BGP Telemetry

Abstract: In network engineering there is a subtle difference between "it works" and "it appears to work". BGP is no exception to this - it is universally deployed building block for a multitude of network services, and yet it may be difficult to get insight into the actual operation of BGP, even more so in the case of possible malicious or unintentional problems. Over time network engineering community has developed practices and tools that would allow for some insight into the operation of BGP, but the state and fragmentation of those practices certainly could be better. IETF has produced a telemetry mechanism for BGP monitoring - BGP Monitoring Protocol (BMP).

This talk will provide a brief overview of BGP monitoring domain problem space, take a quick look at available mechanisms prior to BGP, provide the motivation for the new protocol and the overall concept of routing protocol telemetry, and discuss possible practical use cases of BMP. In addition there will be an overview of the results of BMP implementation interoperability testing that has been ongoing during IETF meetings.

Notes: This has been first presented at MSK-IX peering forum in December 2019. Slides attached. Video recording is here: https://www.youtube.com/embed/ta6BVyBShOY?start=4200&rel=0&autoplay=1 .This is a broad topic, and especially the operational part of the BMP deployment is wide-covering, therefore a longer slot is requested to cover the discussion at the end. There are a few open topic in BMP evolution and a feedback from operational community would be of significant interest. 20 minutes of presentation time and 10 minutes of discussion time at the end would be reasonable time estimate.

Title: Peering Telemetry

Abstract: Data plane telemetry has been around for a while and enjoys a widespread deployment, as well as being considered as an important source of information for capacity planning and anomaly mitigation. There is one large area of visibility that data plane telemetry lacks though – that is correlation of IX control plane events with data plane events and statistics. Having an insight into such correlation would allow for more deterministic and specific reactions to anomalies and events happening in IX environment.

The overall concept is rather simple – take IPFIX and BMP feeds and glue them together in time domain, and then let external entities slice and analyze resulting data.

Notes: This is at a stage of work-in-progress at the moment, both from a prototype perspective, as well as being thought about how this could e made as a usable and replicable framework suitable for a general use. It is expected to be briefly discussed during upcoming LINX 108 meeting.

Primary author

Presentation Materials

Your browser is out of date!

Update your browser to view this website correctly. Update my browser now