Network Analytics @ IETF 115

With 854 onsite and 776 remote attendees, IETF 115 came back to the European continent in London. The picture shows a group of Network Analytics key players and enthusiasts at the wedding

of YANG, the data modelling language for configurational and operational networking data and Confluent Apache Kafka, the de-facto message broker standard. Enabling with such a Data Mesh architecture the future of Close Loop Operated Networks.

At IEPG, were operational relevant topics are being discussed at IETF, and a dedicated side meeting with key network operators, network and analytical vendors and universities, I had the chance to present the challenges in Network Telemetry Data Mesh integration and how we mariage YANG/CBOR with Apache Kafka.

IETF fireworked on Saturday with the hackathon event which attracted 456 engineers across the industry. We participated with the SRv6 data-plane visibility team where we developed and tested the Export of Segment Routing IPv6 Information in IPFIX. We established a multi-vendor SRv6 network topology

where we developed and validated with FD.io VPP, Huawei VRP and pmacct three implementations successfully. The defining key moment was when on one table Yannick Buchs and Severin Dellsperger changed the SRv6 traffic engineering policy, and on another table Benoit Claise and I observed in the Imply Pivot WebUI the change in the forwarding path in real-time. Suddenly realizing that we achieved something which has never been done in history before.

Many thanks to all colleagues from INSA, OST, NTT, Huawei and Swisscom which contributed and helped to make this moment happened.

And here the hackathon presentation and video for the curious.

Besides the tsunami of nine existing IPFIX, BMP and YANG push drafts, we, Paolo Lucente, Camilo Cardona, Alex Huang Feng, Benoit Claise, Jean Quilbeuf, Ahmed Elhassany and Marco Tollini, had a new YANG push and a BMP draft in our baggage.

Alex presented the latest updates on the existing YANG push drafts, where I had the pleasure to introduce a new draft in the NETCONF working group.

• draft-ietf-netconf-udp-notif (UDP-based Transport for Configured Subscriptions) reaching stable state. Enabling JSON and CBOR export at scale with minimal overhead. The udp-notif Library is open-sourced and we have with pmacct a reference implementation.
• draft-ietf-netconf-distributed-notif (Subscription to Distributed Notifications) describing how multiple YANG publishers can export on one node. Enabling that YANG operational metrics can be directly export from network processors similar as IPFIX already does.
• draft-tgraf-netconf-yang-notifications-versioning (Support of Versioning in YANG Notifications Subscription) describes how the YANG push header is being extended with semantic references. Enabling the automated Data Mesh integration without workarounds.

Benoit and me presented at the OPSAWG working group updates on our existing three drafts. For the Export of Segment Routing IPv6 Information in IPFIX draft we asked the working group for IANA early code point allocation.

• draft-ietf-opsawg-ipfix-srv6-srh (Export of Segment Routing IPv6 Information in IPFIX) enables SRv6 data-plane visibility by exposing the SRv6 Segment Routing Header properties in IPFIX. An important aspect for network operators such as Swisscom which transitions their networks from MPLS to IPv6 data-plane. With Huawei and Cisco we have two vendors which verified technical feasibility and work on implementations. Other vendors showed interest as well.
• draft-tgraf-opsawg-ipfix-on-path-telemetry (Export of On-Path Delay in IPFIX) extends IPFIX to export On-Path delay at scale by preserving Flow Aggregation (RFC7015). Enables network operators to monitor delay Service Level Objectives for their customers by having a statistical delay view over their network. That helps to understand to identify where delay in the network for which service and forwarding-path is being accumulated.
• draft-claise-opsawg-collected-data-manifest (A Data Manifest for Contextualized Telemetry Data) working group adoption has been requested. Gives context how YANG push metrics where collected (software version, on-change vs. periodical etc.).

Camilo and Paolo presented updates on the ongoing BMP and a new BMP draft at the GROW working group. The BMP YANG Module draft is now officially adopted at the GROW working group.

Zhuoyao Lin from the école Polytechnique University in Paris contributed BMP high availability to the pmacct open-source project and presented the implementation.

• draft-grow-bmp-tlv (TLV support for BMP Route Monitoring and Peer Down Messages) reached last call and will move on to IESG. This is a prerequisite for draft-cppy-grow-bmp-path-marking-tlv (Path Marking TLV) which brings visibility how BGP paths are being installed into the RIB.
• draft-ietf-grow-bmp-tlv-ebit (Support for Enterprise-specific TLVs) reach working group adoption state. Enabling future development and testing of new BMP applications.
• draft-cptb-grow-bmp-yang (BMP YANG Module). Enabling BMP netconf-yang configuration and exporting BMP statistics with YANG push to monitor the BMP process.
• draft-lucente-grow-bmp-rel (Logging of routing events in BMP) was introduced to GROW. Enabling to expose BGP event, such as route policy drops, in BMP.

We are all looking very forward to IETF 116 in Yokohama Japan where we will be able to meet in person again and engage at the hackathon to extend our SRv6 data-plane visibility capabilities and continue to bring YANG and Data Mesh closer together. Stay tuned!

Thanks to all colleagues who contributed! Feedback and comments as always very welcome.