IETF-99 IPPM Agenda Session
2017-07-19 0930-1200: Athens/Barcelona - Audio stream - IPPM Chatroom
Internet Protocol Performance Metrics (IPPM)
IETF 99 Prague, Czech Republic
Prague Hilton - Room Athens/Barcelona
Wednesday, July 19, 2017
9:30 - 12:00 (noon) CEST (Prague venue local time)
0 - Chair Slides (10 min) - Bill Cerveny / Brian Trammell
Brian Trammell: Chair slides https://www.ietf.org/proceedings/99/slides/slides-99-ippm-chair-slides-02.pdf
RFC 8186 published
draft-ietf-ippm-model-based-metrics on iesg telechat in August 2017
draft-ietf-ippm-6man-pdm-option - Competed iesg evaluation
draft-ietf-ippm-2330-ipv6-01 - Still waiting on 6man
draft-ietf-ippm-alt-mark-05 WGLC complete, waiting for writeup (thanks Carlos!)
draft-ietf-ippm-twamp-yang-03 WGLC starts now - shepherd volunteers on mailer
Readout on agenda - slide :[IPPM Agenda / IETF 99 Prauge]
Working Group Documents
draft-ietf-ippm-[initial,metric]-registry (15 min) - Al Morton
Al Morton: presenting Registry for Performance Metrics
Slide 4: discussion with IANA to arrive at a wide format registry
Slide 5: Additional loss metrics
Slide 6: 3rd section in IANA - extension of metric with mockups. URI to the complete registry description
Slide 7: Net name elements mertric type, units, output types
Slide 8: Name element sub-registries
Slide 9: Question: Does name element cover passive measurement well enough?
Initial Performance Metric Registry Entries
Slide 11:Feedback on registry contents. Fairly easy to put in loss metrics. Simple approach to edit and add registry entries.
Slide 12: New metrics - loss related
Slide 13: Loss metrics added, traceroute metrics proposed,
Registered method for ICMP echo request/reply? Will this be used for basis for comparison?
Will anyone use ping measurements for regulatory comparisons.
Brian Trammell: Broadband mapping project (European) needs this. Use ping metrics for regulatory purpose.
Nudge them to use other measurement metrics (twamp etc) but this may still be needed
Slide 13: passive measurements - TCP measurements transport level passive measurements in use.
Quic focussing on end to end RTT measurements.
Loss measurements is interesting.
Active measurements for latency,loss - less interesting.
Mike Ackerman: Echo request/reply will use it.
Slide 14: DNS loss measuremtns will need some form of ID/sequence
Al: DNS needs sequence number to match q&a and measure.
Nalini Elkins: DNSSec has ID, but that is dropped
draft-ietf-ippm-2330-ipv6 (15 min) - Joachim Fabini
J.Fabini: Update for the IPPM Framework: Adding Support for IPv6 and IP Options
Brian Trammell Slide 6: question about 6lowpan participation?
J.Fabini: Slide 7 on 6lowpan
Brian T: explicit support for 6lowpan - interesting if measurements to compare 6lo to ipv6 traffic, or path comparison measurements
Al: End to end ipv6 and 6lo packets go through a gateway
Spencer Dawkins: AI to take the discussion with the ADs for the 6lo interaction
Slide 8: Is anyone using minimal header? <no one came to mike>
Slide 9: IPv6 ext header treatment
Slide 10: WGLC?
Brian T: not enough reviewers in room. Next rev up will be posted.
Reviewers with IPv6 for shepherding this? <no one in the room>
Mike Ackerman: Should this be carried to 6man for handling for ext headers etc. Is there a need for 6man involvement? Will 6man benefit from this input
Brian T: Will cross post pre-WGLC to 6man
General Working Group Business
Rechartering for IOAM etc (60 min)
Charter Suggestion - Brian Trammell / Bill Cerveny
Brian on Rechartering IPPM slides:
Rename wg -> IPP ‘M’-> s/Metrics/Measurements/
Remove “Specifying network or lower layer OAM mechanisms is out of scope of the IPPM charter. “
This doesn’t effect bringing in IOAM, but will remove the confusion to not do OAM.
Comments on Paragraph 1?
Spencer D: likes the Paragraph 1.
Tal M: Can we get rid of IP in the IPPM? some of the methodologies are not IP specific, e.g., MPLS.
Brian T: Individual comment: May be. But the Type of measurements this wg focusses on IP end point.
Greg M: Likes Tal’s comment. Network and lower layer is confusing, there are overlays. What about the overlays? Overlays add oam and measurements in the overlay layers/sub-layers.
Non-IP OAM can benefit from the methologies defined here.
Brian T: <responding to Greg M as an individual> Key is commonality in methodology across layers and protocols promotes reusability. <Chair hat on> not sure how to add this in the charter.
Applicability of the metrics and measurements in non-IP network - not sure how to scope it.
Greg M: on wire, encapsulation etc can be out of scope in non-IP or overlay networks.
Brian T: OWAMP and TWAMP and anything else IP measurements in scope.
Greg M: BFD as an example.
Brian T: Take to the list a new paragraph for scoping. I(P+)PM?
Tal M: Work being done today already goes beyond IP. We can add text saying that in some cases methologies go beyond IP, or as an exercise drop IP from charter, and see if the outcome makes sense.
Al: Goal is to foster better measurement and commonality across different layers. Make recommendations that are broadly applicable without creating overlapping work.
how about IP Performance Methology ?
Carlos : Great value in Tal’s proposal to go beyond IP. Targeted paragraph to scope the work for non-IP common methodology instead of modifying the name.
Brian T: Is it safe to say if measurement or methodology that are applicable to IP is in scope and making it common/reusable across other non-IP is in scope. While measurement or methodology that are non applicable to IP is not scope.
Carlos Pignataro: Agrees.
Dave ..: Agrees with last comment from Brian T.
Brian T: Any opposition in the room?
ACTION: Will do the wordsmithing on paragraph 1 on the list.
Paragraph 3: dependency on RFC6390
Al: continue to have the reference to RFC6390
Need to be clear about contextual information that will be collected for measurement.
Frank Brockners: Question - there are multiple method, tools, is it possible to qualify confidence of measurements?
Brian T: Method to measure and quantify confidence in a tool/methology is hard.
It's in scope to figure out how to add it in charter - confidence (in measurement Vs the tool/methodology used for measurement)
Brian T: “It may foster this work by defining information and data models for storage and dissemination of measurement data.” came from ANRP workshop.
Carlos Pignataro: Support the comment added. Demarcation b/n work in Lime Vs this needed - for data models and export data. Add more precision to this line.
Brian T: Another Paragraph ,, Carlos P to suggest text.
Frank Brockners: Comments on the same lines - work in Lime Vs scope of doing the data export and modeling in IPPM.
Brian T: Terminology for information and data model in measurement vs operational community is different.
Brian T: Thumbs up from Spencer on the rechartering work.
ACTION: Chairs to make the text edits for charter + text from Carlos and ship it out to Spencer.
draft-brockners-inband-oam-data - Frank Brockners
Addressed comments from Chicago to now -
Sections and text on scope added
Checksum neutral data field update for trace data
Document update from experimental to standards track
3 commercial implmentation announcements and open source implementations of this available
Brian T: Continue to edit the draft as draft-brockers till charter update is done. Consensus on workgroup adoption holds.
Shahram Davari..Broadcom :1. End in-situ OAM. need timestamps in E2E types.
2. PTP to timestamps?
3. Not clear if there is iOAM in the packet or not
Shwetha: On (1) possible to define a new subtype for timestamp. There is a draft already, will send a reference
Frank B: ioam in packet is specific to encapsulation
draft-song-ippm-ioam-scalability - Haoyu Song
draft-amf-ippm-route (15 min) - Jose Ignacio
Greg Mirsky: Have you considered alternate marking methods as one of the hybrid methods to instrument?
Ignacio: Intention is to measure every hop of the flow. With marking method there needs to be something in each hop to collect data.
Brian T: Is a good discussion to have b/n Greg M and Ignacio offline and report in Singapore
Greg M: challenge with round trip packet path being co-routed (??) (with ECMP algorithm that can make the path asymetric?)
Greg M: May be better to split round trip delay are 2 one way measurements.
Ignacio: Will need support from intermediate hops.
The above work is related to rechartering. Will be considered for adoption post rechartering.
TWAMP-Light and BBF (20 min)
Brian T: Context: BBF port assignment request for test traffic (TR69) BBF control plane and twamp data plane.
draft-mirsky-ippm-twamp-refl-registered-port - Greg Mirsky
IANA Service Name and Transport Protocol Port Number Registry had assigned TCP port 862 for twamp-control service TWAMPControl protocol per RFC 5357
•but it had assigned UDP port 862 for TWAMP-Control as well
• Use this document to re-assign UDP port 862 for TWAMP Test Session-Reflector Receiver port as twamp-test service
draft-morton-ippm-port-twamp-test - Al Morton
- IANA re-allocate UDP ports 861 and 862 as shown below, leaving the TCP port assignments as-is:
Service Port Protocol Description
owamp-control 861 tcp OWAMP-Control [RFC4656]
owamp-test 861 udp OWAMP-Test [RFCXXXX]
twamp-control 862 tcp Two-way Active Measurement Protocol (TWAMP) Control [RFC5357]
twamp-test 862 udp Two-way Active Measurement Protocol (TWAMP) Test [RFCXXXX] where RFCXXXX is this memo when published.
Shahram Davari Broadcom : Lite version doesnt have control protocol?
Greg M: TWAMP uses distributed control plane protocol. And other uses for e.g. SDN controller to drive Twamp lite.
Briant T: Rename as STAMP - simple two way active measurent?
Brian T: Summarize differences draft-mirsky-ippm-twamp-refl-registered-port and draft-morton-ippm-port-twamp-test?
Greg M: OWAMP and active measurement terminology clarification.
Brian T: Can you merge the two drafts and then call for adoption?
Al and Greg - Agree
ACTION: Al Morton and Greg Mirsky to merge the two drafts and then call for adoption.
Humm in favor of adopting merged draft - positive.
draft-mirsky-ippm-twamp-light-yang - Greg Mirsky
Shahram Davari Broadcom : There is a problem measuring average - running or weighted, has anyone done it? Has anyone proposed it?
This is difficult when there are large number of flows.
Greg M: For measuring average alternate marking is good
Al: There is a running average in RTP jitter evaluation.
Lightning Talks (Enforced 5-minute limit, as time permits)
draft-fioccola-ippm-multipoint-alt-mark - Giuseppe Fioccola
Al: Suggest reading RFC 5644 that does some of this
draft-mizrahi-ippm-multiplexed-alternate-marking - Tal Mizrahi
Tal: Recommend reading this draft and providing feedback even if not familiar with alternate marking..
<Session concludes .. no time to get to the below talks>
draft-mizrahi-intarea-packet-timestamps - Tal Mizrahi
draft-bhaprasud-ippm-pm - Sudhin Jacob