#acumos-meeting: Acumos TSC
Meeting started by bryan_att at 15:56:57 UTC
(full logs).
Meeting summary
-
- valderrv (valderrv,
15:57:03)
- Bryan Sullivan (bryan_att,
15:58:23)
- Scott Nicolas intros the agenda (bryan_att, 16:01:00)
- Agenda: welcome, TSC organization, release
roadmaps, communication channels, infra (bryan_att,
16:01:53)
- there are 8 members of the TSC for this first
meeting (bryan_att,
16:02:09)
- (intros of those in the room) (bryan_att,
16:03:33)
- 8 TSC members Ofer Hermoni Amdocs, Jack Murray
AT&T, Wang Yi Baidu, Ying Long Xia Huawei, Pantelis Monogioudis
Nokia, Sachin Janokar TechM, Xialong Zhu Tencent (bryan_att,
16:09:30)
- members of the TSC dialing in: Yinglong Xia,
Huawei (bryan_att,
16:16:24)
- TSC Organization (bryan_att, 16:16:35)
- Scott turns floor over to AT&T and TechM to
discuss seed code contribution (bryan_att,
16:17:06)
- Jack Murray : AT&T and TechM have made a
seed code contribution, now in repos and the wiki, along with tools
for building the project, and have deployed an initial root node for
the community at market.acumos.org. AT&T and TechM have shared
an initial set of models there to illustrate model publication and
federation. The contributions are Apache-2.0 licensed. (bryan_att,
16:19:04)
- ... the Acumos day yesterday was intended to
intro the attendees to the community and goals for Acumos. Many
members are expected to leverage such solutions internally across a
broad set of applications. (bryan_att,
16:20:35)
- Sachin Janokar: our vision for Acumos was a
collaborative way to solve the need for ML-based solutions.
(bryan_att,
16:22:54)
- Jack: a couple of things to share in kicking
off: (1) we need an approach of at least two releases per year (a
stable release branch from what is a continually developed
master). (bryan_att,
16:23:59)
- ... the first release is proposed to be called
"1.0". What we name these is something to be discussed, e.g. greek
gods (but not cities). (bryan_att,
16:25:01)
- ... (2) the overall TSC organization, shown in
a word document shared by Jack with a proposal of sub-committees and
projects (bryan_att,
16:28:27)
- ... projects: Integration and Test (Test Use
Cases, Release Management, CI/CD process) (bryan_att,
16:30:25)
- Q on the catalog growth re the role of the
Marketing committee (or the LFDL board): answer as proposed is a
collaboration of the TSC and BOD committees (bryan_att,
16:31:53)
- Raj: where will owners of Model contributions
be tracked? (bryan_att,
16:32:43)
- Jack: under the community; the architecture
committee could track use cases and associate to models contributed
by members (bryan_att,
16:33:40)
- Raj: I see one track as enhancing the platform;
another being modeler communities that want to just onboard their
models (bryan_att,
16:34:39)
- : Jack I see the proposed Acumos.org driving
both of those goals (bryan_att,
16:35:15)
- projects (continuing): Acumos Portal and UX
(end-to-end use csaes, roles, search, ratings) (bryan_att,
16:36:44)
- Catalog, Data Model, Data Management
(federation, validation and compliance) (bryan_att,
16:37:39)
- ... Training (Data Management, Data Handling,
Acceleration, Cloud) (bryan_att,
16:38:32)
- Design Studio (Microservices) (bryan_att,
16:38:48)
- Deployment (local, cloud) (bryan_att,
16:39:04)
- Common Services (new) (bryan_att,
16:39:15)
- ... Model Onboarding (new modeling methods,
optimizations) (bryan_att,
16:39:58)
- ... this is related to but a different spin to
the proposal on the wiki that Manoop and Bryan worked on:
https://wiki.acumos.org/display/AC/Acumos+Architecture+Mapping+to+Repos+Jira+etc
(bryan_att,
16:40:53)
- Ofer: maybe data management better called "data
set management" (bryan_att,
16:41:58)
- Jack: Data Management along with Catalog etc
was meant to capture the ability to contribute data sets to the
catalog, e.g. for training (bryan_att,
16:42:44)
- ... the main goals of this proposal is to
result in a small number of standalone projects (bryan_att,
16:43:46)
- ... training should be a primary area for
Acumos, though very much a work in progress (bryan_att,
16:44:19)
- (projects continued): Documentation
(bryan_att,
16:45:57)
- Common Services (microservice generation,
download, import.export) (bryan_att,
16:46:32)
- License Management (bryan_att,
16:46:45)
- OAM&P (installation, logging,
metrics) (bryan_att,
16:47:05)
- Paul: is license management about software, the
models, or both (bryan_att,
16:47:30)
- Jack: the models (bryan_att,
16:48:00)
- Nat: the platform is Apache 2.0 (bryan_att,
16:48:23)
- Jack: the platform is Apache 2.0; we have to
consider how to address other software we need to integrate
with (bryan_att,
16:48:59)
- https://zoom.us/j/137231283
(bryan_att,
16:49:11)
- (? ZTE) where are the datasets from and how
associated with companies (bryan_att,
16:50:28)
- Jack: Datasets fall under those associated with
a model, and those contributed to broader work. Some datasets may be
referenced, others contributed ala software. LF has indicated a
different project re establishing guidellines for dataset
sharing (bryan_att,
16:51:32)
- Scott: are considering two licenses focused on
data; one a Apache 2.0 and another a copyleft based license. Do we
anticipate a lot of data contributions> (bryan_att,
16:52:23)
- Jack: we expect data associated with models,
but not all of that will be public data (bryan_att,
16:52:56)
- Pantellis: we expect there to be data available
of sufficient size to effectively test the platform (bryan_att,
16:53:34)
- Jack: we expect the databroker to enable
connecting models to external data, referenced for training purposes
or as an example (bryan_att,
16:54:17)
- Raj: the availability of training datasets will
be important (bryan_att,
16:54:42)
- Jack: we want to promote all model contributors
to provide training datasets, not necessarily large but
adequate (bryan_att,
16:55:21)
- ... that said we need to manage the amount of
data being replicated across portals (bryan_att,
16:55:47)
- (Q on the phone): if I have a model to
contribute how do I get that into the marketplace? (bryan_att,
16:56:20)
- Jack: the federation capability of the platform
is intended to support that, as long as systems implement the
federation API e.g. as an Acumos portal instance or some other
Acumos compliant system (bryan_att,
16:57:22)
- (Q on the phone): If I have an internal
marketplace I may want to use an Acumos instance as a gateway to
publish models into the ecosystem (bryan_att,
16:58:12)
- Sachin: what about runtime envs to allow devs
to download and use models locally? (bryan_att,
16:58:33)
- Jack: we will support public clouds and
internal openstack/k8s etc cloud envs as a launch target
(bryan_att,
16:59:17)
- ... other enhancements might be one click setup
of toolkit-specific test/training envs on a workstation for
example (bryan_att,
17:00:03)
- ... training envs will likely provide some
local options, e.g. a GPU cluster dedicated to training (bryan_att,
17:00:39)
- Nagib (Intel): how can Q's get answered about
the platform (bryan_att,
17:01:26)
- Jack: stack overflow, Jira, IRC etc. These are
the common channels used in open source. (bryan_att,
17:02:21)
- ... mail lists etc (bryan_att,
17:02:55)
- new mail lists are setup and accessible thru
the Acumos website (bryan_att,
17:04:02)
- Bryan: See
https://wiki.acumos.org/display/INFRA/Project+Infrastructure for a
full list of the communication and collaboration tools (bryan_att,
17:06:30)
- Jim (IBM): is this the list for Acumos or LFDL?
what is the process for adding projects to LFDL? (bryan_att,
17:07:08)
- Jack: talk to Scott about new projects under
LFDL, and shared committees e.g. BOD, TAC, Marketing, etc
(bryan_att,
17:07:40)
- Jack: critical success factor will be
federation of existing marketplaces into Acumos (bryan_att,
17:08:03)
- Jack: at the booth we will show a demo of
federating with an ONAP instance Catalog service (bryan_att,
17:08:38)
- Jack: other suggestions? (bryan_att,
17:09:19)
- Pantelis: suggest the OAM and Integration and
Test (bryan_att,
17:09:35)
- ... be integrated? (bryan_att,
17:09:46)
- Jack: they are already a big project
(bryan_att,
17:10:05)
- Bryan: putting these into a single project
would mix roles and technical focus - they are probably better
separate (bryan_att,
17:11:52)
- Jack: AT&T will probably contribute some
lab space for testing (bryan_att,
17:12:36)
- Bryan: also the TSC should consider the
need/approach to a community lab program ala the OPNFV Pharos
program: https://wiki.opnfv.org/display/pharos/Community+Labs
(bryan_att,
17:13:37)
- Jack: back to Scott, how do we position for
success? (bryan_att,
17:13:56)
- Chris: meeting cadence assumed to be
weekly? (bryan_att,
17:14:13)
- Jack: maybe initially weekly, then evolving to
a TSC alternating with projects,weekly (bryan_att,
17:14:58)
- Jack: we need to call for votes on TSC chair,
release roadmap, committees, etc (bryan_att,
17:15:43)
- Scott: TSC chair election process would be run
by email with a nomination period, statement of intent and bio, and
run the vote pretty quickly. I propose a nomination period opening
now and ending next Tuesday, with voting thru next Friday.
(bryan_att,
17:18:16)
- Ofer: what other positions should we
consider? (bryan_att,
17:19:24)
- Jack: any need for a co-chair position?
(bryan_att,
17:20:16)
- Sachin: how do we form the committees?
(bryan_att,
17:20:47)
- Jack: we vote on the committees and let them
form (bryan_att,
17:21:01)
- Susan: Yes - so is the other location reserved?
should we expect to see the code at https://github.com/acumos
sometime? (bryan_att,
17:22:36)
- Scott: calls for TSC agreement on the process
for TSC Chair election (bryan_att,
17:23:56)
- The TSC agreed with the proposed process for
TSC Chair election (bryan_att,
17:24:17)
- Scott: should the project approve the project
structure Jack has proposed? (bryan_att,
17:25:40)
- The TSC agrees to the project structure as
proposed by Jack (bryan_att,
17:26:07)
- Scott: should the project implement a bi-annual
release schedule? (bryan_att,
17:26:45)
- The TSC agrees a bi-annual release
schedule (bryan_att,
17:29:06)
- Bryan: shows the wiki page
https://wiki.acumos.org/display/AC/Acumos+Architecture+Mapping+to+Repos+Jira+etc
where the association of projects to repos, and links to where to
find where who are the committers etc (bryan_att,
17:39:29)
- Jack: suggest that those who want to contribute
to the committees reach out to Annie to express interest and get
connected to the mail list etc where the committees will
collaborate (bryan_att,
17:41:04)
- Bryan: suggests we have (1) TSC meetings; (2)
project meetings; (3) Community-wide meetings ala the Model that has
worked well in OPNFV (bryan_att,
17:44:26)
- Pantelis: asks for clarification of use of
Slack (bryan_att,
17:46:26)
- Jack: it seems that we would incur some
additional costs now for Slack integration with IRC etc, so it
appears we will better focus direct communication on IRC
(bryan_att,
17:47:22)
- Pantelis: what about integration with Github
etc for notifications of issues, etc (bryan_att,
17:47:58)
- Bryan: github will be a mirror, but generally I
don't think we would expect it to be a way that issues etc are
brought in, if we focus users and others on JIRA or other avenues
e.g. Stackoverflow for input. Also there is strong integration with
Git/Gerrit and IRC that should provide needed notifications.
(bryan_att,
17:49:41)
- Scott: the TSC will meet in 2 weeks at April
9th 6AM Pacific time. (bryan_att,
17:51:02)
- Bryan: we will post these minutes and future
TSC meeting schedule, agendas, etc on a Wiki space/page created for
that, and send out the links to the lists etc. (bryan_att,
17:51:54)
Meeting ended at 17:52:00 UTC
(full logs).
Action items
- (none)
People present (lines said)
- bryan_att (108)
- collabot` (4)
- ChrisPriceAB (2)
- valderrv (2)
- Guest83150 (1)
Generated by MeetBot 0.1.4.