Tuesday 19th March 2024, 09:30-11:30 AEST
Meeting room M1, https://meetings.conf.meetecho.com/ietf119/?group=ippm
Chairs: Tommy Pauly & Marcus Ihlar
Note taker: Stuart Cheshire
Chairs opened meeting with introduction.
IPPM has moved from the Transport Area (tsv) to the Operations and
Management Area (ops).
Gregory Mirsky (Ericsson): Joining the Ops area makes a lot of sense.
draft-elkins-ippm-encrypted-pdmv2
Michael Ackermann (Blue Cross) presented brief update on document
status.
Bjørn Ivar Teigen (Domos) gave a review of draft-olden-ippm-qoo-02.
Proposal to add seperate NRP (Network Requirements for Perfection) and
NRPoU (Network Requirement points of uselessness) for packet loss.
Goal is to express requirements of different applications in a uniform
way.
Tommy Pauly (Apple): Is the scale from “perfect” to “unusable” linear?
Something else? Is there a way to express that?
Bjørn Ivar Teigen: For now we are assuming linear. There is a trade-off
between simplicity and precision.
Martin Duke (Google): You certainly can model packet loss as (infinite)
latency.
Etuate Cocker (Exclusive Networks): Different apps behave differently.
Voice (telephony) apps have different requirements from Facebook.
Bjørn Ivar Teigen: This is why different apps have different NRP and
NRPoU.
Luis Contreras (Telefonica Innovacion Digital): Applications can adapt
to different network conditions, rather than assuming an unvarying
requirement of the network.
Bjørn Ivar Teigen: Some applications are better at adapting than others.
draft-fz-ippm-alt-mark-deployment
Giuseppe Fioccola (Huawei Technologies) presenting.
No comments from the room.
draft-ietf-ippm-ioam-data-integrity
Justin Iurman (University of Liege) presenting.
Gregory Mirsky (Ericsson): DEX does not necessarily require generating a
packet.
Gregory Mirsky (Ericsson): We need to distinguish different IOAM trace
options. The proof-of-concept was primarily developed by the Service
Function Chaining group, but that is no longer progressing. Is this of
interest to anyone?
Tommy Pauly (Apple): Thanks for doing this analysis. I agree that option
2 makes the most sense. What is the implementation status of this? Does
using IPSEC change the nature of what’s being measured? We would love to
see feedback from other people.
Tommy Pauly (Apple): Proposed moving ahead with only Option 2 for IOAM
integrity?
Yes: 3
No: 1
No opinion: 9
Paolo Volpato (Huawei): I would like to see more flexibility than just
option 2.
draft-gandhi-ippm-stamp-ext-hdr
Rakesh Gandhi (Cisco) presenting.
Xiao Min (ZTE Corporation): Would you consider adding a new STAMP TLV to
reflect the IPv6 header?
Rakesh Gandhi: That seems like a good idea.
Giuseppe Fioccola (Huawei Technologies): We could combine this with the
similar document defining the hop-by-hop extension.
Gregory Mirsky (Ericsson): This draft is well written and ready for
working group adoption.
draft-spiegel-ippm-ioam-rawexport
Justin Iurman (University of Liege) presenting.
Thomas Graf (Swisscom): Thank you for this work. We need to define the
exporting protocols.
Gregory Mirsky (Ericsson): In the document different trace options are
considered differently. Would they result in a different export?
Justin Iurman: I don’t remember.
Thomas Graf (Swisscom): It seems we could decompose the IOM options into
IPFIX.
Martin Duke (Google): How is this different from direct export?
Justin Iurman: This draft defines the say to export the data.
draft-mirsky-ippm-asymmetrical-pkts
Gregory Mirsky (Ericsson) presenting.
Etuate Cocker (Exclusive Networks): Can you say more about the rate
limiting?
Gregory Mirsky: Setting the number of reflected packets to zero
effectively creates a unidirectional measurement. When both TLVs are
present and contradict each other we need to specify what to do.
Bjørn Ivar Teigen (Domos): I find the proposal interesting and think it
is worth. We need to be careful about the case where one packet can
elicit multiple reflected packets.
Rakesh Gandhi (Cisco): Slide 4 says, “A Session-Sender SHOULD NOT send
the next STAMP test packet…” What should it do instead?
Li Zhang (Huawei) presenting.
Thomas Graf (Swisscom): Clone a packet based on an IOM flag. A telemetry
protocol should not be intrusive.
Gregory Mirsky (Ericsson): I agree with Thomas. Flooding the network
might not be a practical approach.
Giuseppe Fioccola (Huawei Technologies) presenting.
draft-he-ippm-integrating-am-into-ioam
Xiaoming He (China Telecom) presenting.
Gregory Mirsky (Ericsson): I wonder about the motivation for this work.
Work on alternative marking in IPv6 is proceeding.
Thomas Graf (Swisscom): Section 3 in the document mentions problems with
excessive amounts of exported data.
draft-he-ippm-extensions-incorporating-am
Xiaoming He (China Telecom) presenting.
draft-song-ippm-ioam-ipv6-support
Haoyu Song (Futurewei) presenting.
Thomas Graf (Swisscom): Thank you for raising this point. I would like
to hear from other vendors.
Gregory Mirsky (Ericsson): Thank you for your analysis. We already have
a solution for the scaling problem: Direct expert.
Thomas Graf (Swisscom): I think this is solvable through IOAM direct
export.
Hongyi Huang (Huawei) presenting.
Marcus Ihlar: You should take a look at the Quality of Outcome document
we have adopted.