[alto] ALTO Minutes from IETF-108

[Jan Seedorf <[e-mail]>]

Dear all,

please find below the minutes from the ALTO session at IETF-108 which I 
just uploaded to the datatracker. Please let us know any comments or 
corrections you might have.

A big thanks to Qiao and Sabine for taking the notes!

  - Jan & Vijay


****

----- AGENDA & MATERIAL -------------------------------------------

IETF 108, Virtual Meeting
ALTO WG, Mon Jul 27, 2020 14:10 - 15:50

14:10 - 14:15   5"   Chair slides
14:16 - 14:26  10"   Unified Properties, Sabine Randriamasy [1]
14:27 - 14:35   8"   Performance metrics, Richard Yang [2]
14:36 - 14:41   5"   CDNI, Jensen Zhang [3]
14:42 - 14:52  10"   Path vector, Kai Gao [4]
14:53 - 15:50  57"   Recharter discussion

[1] https://tools.ietf.org/html/draft-ietf-alto-unified-props-new-12
[2] https://datatracker.ietf.org/doc/draft-ietf-alto-performance-metrics/
[3] https://tools.ietf.org/html/draft-ietf-alto-cdni-request-routing-alto-12
[4] https://datatracker.ietf.org/doc/draft-ietf-alto-path-vector/

Slides
https://datatracker.ietf.org/meeting/108/session/alto

----- ATTENDANCE: 40-43 people -------------------------------------------
Bluesheet: 
https://www.ietf.org/proceedings/108/bluesheets/bluesheets-108-alto-202007271410-00.txt 

WG Chairs: Vijay Gurbani, Jan Seedorf
Area Director: Martin Duke

Minute takers: Sabine Randriamasy, Qiao Xiang

----- ACRONYMS AND NAMES USED IN THE MINUTES 
-------------------------------------------
CX Chunshan Xiong
DL Danny Lachos
JS Jan Seedorf
JZ Jensen Zhang
KG Kai Gao
MD Martin Duke
RY Richard Yang
SR Sabine Randriamasy
VG Vijay Gurbani

APM     ALTO Performance Cost Metrics (WG draft)
CDNI    Content Delivery Network Interconnection (CDNI) Request Routing: 
CDNI
                      Footprint and Capabilities Advertisement using 
ALTO (WG draft)
PV     Path Vector (WG draft)
UP     Unified Properties for the ALTO protocol  (WG draft)
EG     e.g.

----- MINUTES -------------------------------------------

----- 0. Chair slides

Overall document status: cost calendar and SSE are in RFC editor. CDNI 
has passed WGLC. Unified properties is in WGLC.

The focus of the WG is to finalize the charter items. Right now things 
are going well. unified properties is in WGLC, and the chairs things 
path vector and performance metrics are also ready for entering WGLC. 
However, due to their dependency on unified properties, the chairs and 
the AD decide to hold them from WGLC until unified properties passes WGLC.

In addition, this meeting will also discuss rechartering. The chairs are 
happy to see the discussion and energy on the mailing list for 
rechartering items.
VG on the ALTO rechartering disc kick-off. The goal is to get a sense, 
as meant in RFC2418, (see slide 7: does charter address relevant issues 
for the Internet community? Are goals specific and achievable (in 
reasonable time) ? Does the WG have the expertise ? ) Goal is not to 
take a re-charter decision during IETF 108 but open the discussion.  The 
proposed stuff should be achievable. So expertise and energy is needed in WG


----- 1. Unified Properties:
Sabine introduces the key design update and text update in the draft. In 
particular, the definition of "defining information resource" is 
introduced. The draft is in WGLC until August 7.

Vijay: the draft is fairly mature.
Jan: people are encouraged to read the document and if there is any 
question, please send to auhors. SR adds 2 other drafts depend on UP so 
UP needs to be clear.


----- 2. Performance Metrics:

Richard introduces the key design updates from v10 to v12 to address key 
comments from 107 meeting: (1) how to choose types of metrics => ALTO 
provides guidance, not measurement framework, and 4 types of guidance 
are chosen; (2) how to conform to rfc6390 => define metrics in the 
document by defining metric name, metric description, units of 
measurment and measurement points, and provide context information for 
coarse-grained information such as routing system information; (3) how 
to handle different statistics => provide complete statistic by adding 
stddev and stdvar and (4) how to convey freshness => use HTTP Last-Modified.

Vijay: after the minor edits suggested by the authors, this document is 
ready for WGLC after UP.


----- 3. CDNI:

Jensen presents the key changes from v11 to v12: (1) generalize media 
types and attribute names to make FCI a special case, and (2) update 
sections related to unified properties, such as representing footprint 
objects as property map entities and representing CDNI capabilities as 
property map entity properties.

Vijay: this document also depends on unified properties.
No questions so this draft will move ahead.


----- 4. Path vector:

Kai presents the updates and status of this draft. This draft is waiting 
to start WGLC, depending on unified properties. Key updates in v11 
includes text updates on the motivation example of flow scheduling, the 
discussion on ephemeral ANE and persistent ANE, and to be in sync with 
the unified properties draft. The presented design updates now support 
non-ephemeral ANEs, as it was the case before, and extend the 
application scope initially restricted to flow scheduling, to paths 
including persistent ANEs.

Vijay: PV is a major deliverable of the first recharter, and should 
start WGLC after or in parallel (TBD) with unified properties. If it is 
ready, the chairs will issue a WGLC.


----- Recharter discussion:
- Slides: Overview
https://www.ietf.org/proceedings/108/slides/slides-108-alto-alto-re-charter-overview-00

Richard presents the recharter overview slides. The current ALTO 
high-level goals is to provide network information to application in an 
abstraction that simplify complex network information. Its bigger 
context is network-application integration. The current ALTO framework 
and its deployment is reviewed.

Following the guidances of AD (Martin), i.e., relevance, review and 
feasibility, the discussion on recharter is organised based on use cases 
(cellular, mEC, huge data, automation, interdomain, ALTO extension) and 
structure (abstractions, transport services, backend and 
server-to-server). A total of 16 1-page presentations are prepared, but 
only 7 are presented due to the time.

- Slides: 1Slider for proposed extensions
https://www.ietf.org/proceedings/108/slides/slides-108-alto-one-slide-alto-extensions-for-recharter-discussion-03

The 7 presented ones are:

(1) (Tencent and China Mobile) MoWIE: network information exposure 
framework for cellular networks (e.g., ABR for cloud gaming, video QoE 
prediction, and TCP perform optimization);
(2) (Telefonica) Using ALTO to determine service edge;
(3) (T-Mobile) ETSI ZSM use case
(4) (Tsinghua U and Shenzhen U): Deliver functions to edge over ALTO;
     Richard comments that CDNI is a good starting point for this item.
(5) (ESNet, Caltech, CERN): ALTO extensions for flexible resource 
information and interaction models are needed to support huge data in 
the SENSE project;
(6) (Telefonica) ALTO extension to support BGP communities
(7) (UNICAMP) Multidomain ALTO services
(12) (Tencent / China Mobile) In-band and out-of-band network 
information exposure
(16) (T-Mobile) ALTO automation extension

The 9 unpresented ones are:
(8) (Tongji U/Sichuan U) Predictive throughput for TCP reactive flows
(9) (Tongji U) Flow-based network information query
(10) (Nokia) ALTO services extension supporting cost context
(11) (Yale) Generic query language extension for ALTO
(13) (Yale) ALTO transport extension supporting HTTP/2
(14) (Tongji U) ALTO transport extension with multipart
(15) (Telefonica) Automatic derivation of ALTO information from southbound


--- Discussion:

Vijay: generic query extension seems interesting.
Andre Bondi (chat channel): I think it would be useful to describe how 
metrics tie back to use cases and how they inform us about network usage 
and use case impact.

Vijay said we should open-up the floor to other ideas. Future meeting 
should be purely on re chartering.
Martin Duke objected that we cannot take 16 drafts as milestones.
Vijay/Jan/Martin agree on limiting to 3-5 re-chartering items.

Vijay/Jan: Instead of waiting for next IETF, we need an interim meeting 
to tie down 3-5 deliverables?

Lyle Bertz: Is this correct? 1. narrow down work (combine by themes) 2. 
try to recharter prior to the next IETF meeting.

Martin: The next deliverable is a draft charter, which requires some 
downsize select and consensus gathering.

Vijay and other attendants agree an interim meeting would be helpful.

****