OPSAWG meeting

IETF 113
Chairs: Tianran Zhou, Joe Clarke, Henk Birkholz
Eliot Lear as scribe.

Joe brought the meeting together at 14:31 CET.
Note well et al presented.

OpsAWG Section

Working group status presented.
Agenda was bashed (no bashing)

A Network YANG Model for Service Attachment Points

Mohamed BOUCADAIR
https://datatracker.ietf.org/doc/draft-ietf-opsawg-sap/
Reviewed work from IETF 112.
Presented some issues with the design.
A need to display multiple SAPs per services.
Need to associate between service attachment and physical attachments.
Spec has been revved.
- Now structured based on the service itself.
- Lots more examples, as to how to manipulate the model.

Scott Mansfield: is service type really a type or a pointer to an instance of a service? You can only have one service for a service type for a node? Explain

Med: you can have multiple services under the same node. You can demultiplex off (didn't catch)
Rob Wilton: Wouldn't you end up binding the sap against an sub-interface?
Med: not necessarily-pleaes clarify
Rob: If you have multiple saps how are they modeled?
Med: I will check against sub-interfaces offline.
Dhruv Dhody: There is a document in CCAMP, and they are asking where to do the network inventory?
Med: it is an inventory model, but up to the WG to check other work.

Next steps:

Only nits from Yang Doctors
Ready for WGLC?

A YANG Model for Network and VPN Service Performance Monitoring

Bo Wu
https://datatracker.ietf.org/doc/draft-ietf-opsawg-yang-vpn-service-pm/

WGLC LC is closed.
One major change suggested: vpn-pm-type.
Add a new choice to cover 2 VPN PM types.
Joe Clarke: do you have enough input from the group and others?
Bo: we published revision 4 that already adds this choice. Do we need another last call?
Joe: change might be substantial, might need another LC.
Bo: that's okay with me.

SAIN Work

Benoit Claise
https://datatracker.ietf.org/doc/draft-ietf-opsawg-service-assurance-yang
https://datatracker.ietf.org/doc/draft-ietf-opsawg-service-assurance-architecture

Please see earlier intros for an intro.

Latest version adds an example of avoiding circular depeendencies.

Needed a way to report "missing" via a YANG union.

Needs one minor update for a counter.

Joe: ???

Then onward to WGLC.

Transport Layer Security Verion 1.3 (TLS 1.3) Transport Model for the Simple Network Management Protocol Version 3 (SNMPv3)

Ken Vaughn
https://datatracker.ietf.org/doc/draft-ietf-opsawg-tlstm-update/

Big change: how to handle hashing table?
Revise the definition of fingerprint algorithm.
A number of constraints added.
CAPITALIZATION OF KEYWORDS

Next steps

More work needed (see slide 5 in preso)
Hoping to be close to WGLC soon after.
Henk: COSI is working on a hashing algorithm index. Please review.
Joe: be careful about the registry name, and maybe there is overlap.
Kenneth: we want to avoid changes to TLS 1.2 tables.
Rob: if new items were added to TLS 1.2, what process would be followed to add into this registry? Is it meant to be a superset of 1.2?

Joe: We don't expect more algorithms.

Rob: so this registry is just seeded by TLS 1.2?
Eliot: this depends on registry policy. Someone should update that policy.

Non-adopted work

Export of Segment Routing IPv6 Information in IP Flow Information Export (IPFIX)

Thomas Graf
https://datatracker.ietf.org/doc/draft-tgraf-opsawg-ipfix-srv6-srh/

Goal: bring dataplane visibility to SRv6.
Describes format for SRH info in IPFIX. Supports C-SID.

Eliot: any implementations?
Thomas: hoping for code by IETF 115.

Joe: We will do CFA on list.

Data Model for Lifecycle Management and Operations

Marisol Palmero
https://datatracker.ietf.org/doc/draft-palmero-opsawg-dmlmo/

Good number of changes since last version.
An example demonstrated.
Would like more feedback. Lots of collaboration going on.
Wnat to evolving YANG models
Virtual meetings taking place every two weeks.

Rob Wilton: interesting problem to tackle. Tricky to make sure it has enough flexibility for different vendors but also enough commonality.
OpenConfig is trying to similar stuff; use identities rather than enumerations.

Bo Wu: Is this a device model or a northbound interface model froma controller?

Marisol: it's an asset model.
Benoit: MULTIPLE DEFINITIONS OF INVENTORY!!. We need some reconciliation on this matter. Maybe a grouping in YANG?
Benoit: Yes, licensing is very complex, especially for multivendor- may be too complex.

14/20 found work interesting.

Marisol: still want to delve through the use cases, to see how far we can get.

+1 from Rob to Eliot saying: get IETF confirmation that there will be no new TLS 1.2 registry entries.

Data Manifest for Streaming Telemetry

Benoit Claise
https://datatracker.ietf.org/doc/draft-claise-opsawg-collected-data-manifest

Porposes two manifests:
- platform menaifest
- data collection manifest

Next steps:
- A few improvements needed (integrity, provenance)
- a few open questions

WG adoption?

Rob Wilton: I recognize the problem statement. You are returning meta-data per subscription. Do you need the meta data being pushed off the box. per leaf node leaves you with too much data. Good to work on, but tricky problem.

Benoit: agree.

Rob: good to know when a value was written and how fresh it is.

Henk: lots of meta data can change. Interesting problem.

Diego Lopez: we are feeling the deluge. This is needed.

Inband Flow Learning Framework

Minxue Wang
https://datatracker.ietf.org/doc/draft-hwy-opsawg-ifl-framework/
lots of discussion around a framework that allows for adaptive telemtry management

Eliot: needs more info to understand the details.
Joe: might be some flows that may be predetermined?
Minxue: customer or operator could initiate a flow; but there may be some privacy considerations in some scenarios.

Rob: looks like it overlaps with IPPM: you may wish to present there. Need a better feel for the scope.

Problem Statement and Use Cases of Adaptive Traffic Data Collection

Xiaoming He
https://datatracker.ietf.org/doc/draft-he-adaptive-collecting-problem-usecases/

Problem: need granular information during bursts that cause problems.

Eliot: the EVENT-MIB allows for threashholding that might help. Also, you may wish to look at the ANRP presentation on A/B bias avoidance.

YANG model for Data Export over IP Flow Information Export (IPFIX) Protocol

Anand Arokiaraj
https://datatracker.ietf.org/doc/draft-arokiarajseda-ipfix-data-export-yang-model/

Pain: SCTP is not well supported, lots of outdated references.
This is a refresh, just focusing on the bulk data export model.
Rob: thank you for perservering. I'd prefer that this go through the WG. Mostly just YANG. Could the WG give it some attention.
Eliot: Maybe give a little time to read?

Ops-Area Section

Administrivia - scribes, minutes, etc.
Warren / Rob

Rob: Hello in person! We'd be happy to have some more volunteers.

Warren: Need some candidates for Ops side of Ops AD.

Open Mic

Al Morton: quic manageability wanted to recommend network admission settings where we would not look at version # for admission control. But that's enforcement of policy. That strikes me as out of bounds.

Eliot: what was their logic?

Al: version numbers will change fast, but the underlying sentiment was that this might ossify networks. It was a view from the past.

Warren: you may be doing admission control on something arbitrary. We try to resolve this through our reviews, but we can raise a DISCUSS.

Al: does this concern other operators?

Rob: Docs need IETF consensus not just WG consensus. Could this also be reworded to go into explanations?

Al: tried that (and many other things)

Diego: Gory wrote a draft about operational implications of pervasive encryption. But this group might have more views about this.