#acumos-meeting: Acumos TSC Community (Product) Committee
Meeting started by bryan_att at 17:17:59 UTC
(full logs).
Meeting summary
- Process for requirements (bryan_att, 17:18:58)
- Jamil: recommend requirements (features etc)
come from the community (product) committee (bryan_att,
17:19:56)
- Nat: agree (bryan_att,
17:20:11)
- strawman of requirements are on the etherpad at
https://etherpad.acumos.org/p/release-plan-athena (bryan_att,
17:21:45)
- Jack: recommend that M1 be focus of the
community committee; as of M2 the arch committee can do the detailed
planning (bryan_att,
17:22:31)
- Bryan: see the existing role/process input on
the wiki https://wiki.acumos.org/display/ACCOM and
https://wiki.acumos.org/display/AR (bryan_att,
17:23:59)
- Ofer: skeptical at this point for the role of
the community committee for R1 - for R2 it's more feasible to do M1
planning (bryan_att,
17:25:10)
- Jack: we should still have discussion about
what can be done until Sept (bryan_att,
17:26:23)
- Jack: we should be able to document what will
be possible in the release as of Oct (bryan_att,
17:26:56)
- Jamil: Nokia (Pantelis) presented an
orchestrator use case (bryan_att,
17:27:20)
- Ofer: that was discussed in this meeting
earlier (bryan_att,
17:27:56)
- Kazi: how to define orchestrator (bryan_att,
17:28:50)
- Bryan: as Pantelist described it, the
orchstrator manages the lifecycle for the model from deployment
through its entire production life\ (bryan_att,
17:30:23)
- Jack: we will need to address the lifecycle of
models vs predictors (bryan_att,
17:31:26)
- Jamil: we can talk about use cases for ML/AI
here, not just for the platform functionality (bryan_att,
17:31:45)
- Jamil: recommend we discuss 2-3 use cases for
the platform in this committee (bryan_att,
17:32:39)
- Bryan: that is in scope of this committee
AFAIU (bryan_att,
17:33:16)
- Jack: we need to determine for the release to
be successful, what does it need to do/support to meet the success
criteria, e.g. # of models in the marketplace, # of active
organizations, etc (bryan_att,
17:34:09)
- Bryan: where are the success criteria defined
? (bryan_att,
17:35:20)
- Jack: they need to be related to the proposed
features as addressing some larger goal (bryan_att,
17:35:44)
- Bryan: so where are those larger goals
defined? (bryan_att,
17:36:25)
- Ofer: we started to consider such things with
Pantelis' input paper about orchestration (bryan_att,
17:37:10)
- Jack: what are people wanting from the current
platform is what we need to determine (bryan_att,
17:39:20)
- Jack: this is essentially a product
requirements spec, e.g. what Acumos is (a place to onboard and share
models), and a way to increase use/reuse of models by the
community (bryan_att,
17:40:29)
- Jack: what is the #1 thing we need to do to
grow the community? make it easy to onboard, find models, motivate
people to contribute here vs somewhere else (bryan_att,
17:41:15)
- Jack: we need to kick off the licensing work
though it will not be complete for R1; we can release a spec for
licensing at least (bryan_att,
17:41:56)
- Jack: not all deliverables need to be finished
software (bryan_att,
17:42:11)
- Jack: also we need to improve lifecycle
mgmt (bryan_att,
17:42:51)
- Jack: also how to safeguard model sharing
(protect from malware, unclear provenance, ...) (bryan_att,
17:43:19)
- Jack: we currently have a validation function
in the seed code (bryan_att,
17:43:48)
- Bryan: there is current content re validation
goals on the wiki, e.g. see
https://wiki.acumos.org/display/SEC/Security+Scanning on the
Security committee page
https://wiki.acumos.org/display/SEC/Security+Home (bryan_att,
17:46:07)
- Jamil: also we need to consider model
portability, to other standards etc (bryan_att,
17:46:25)
- Jack: if these ideas can be broken down, the
arch team can consider how they can be implemented (bryan_att,
17:47:02)
- Jack: we have till June 18 to complete a
requirements plan (bryan_att,
17:47:29)
- Jack: maybe grouping re M1/2/3 is not ideal;
code freeze as of July 16 gives 1.5 months test time (bryan_att,
17:48:07)
- Jack: how much testing time is needed
post-freeze? we do continual test and intend to increase code
coverage (bryan_att,
17:48:37)
- Anwar: as long as IST and E2E testing have
adequate time, it's ok. We will need a few weeks for E2E
testing. (bryan_att,
17:49:13)
- Jack: code freeze is 7/16 and 9/3 for test
complete, so 1.5 months (bryan_att,
17:49:46)
- Jack: what is in RC0 that takes a month?
(bryan_att,
17:50:08)
- Jamil: to complete system test for the use
cases we agreed to support (bryan_att,
17:50:35)
- Jack: so from 7/16 to 10/8 is all testing
time (bryan_att,
17:50:58)
- Jamil: 7/16 is M3 (API freeze), 9/4 is M4 (code
freeze). RC0 is 10/8 (E2E functional test complete), RC1 could be
for E2E use case testing (bryan_att,
17:52:48)
- Jack: we should document the release plan on
the wiki (bryan_att,
17:53:41)
- Bryan: I will add the latest info from Jamil
into a release page for Athena under the release space (bryan_att,
17:54:21)
- Ofer: from what we discussed today, we can
progress on these calls with enough participation; but skeptical on
what we can get completed for the 1st release (bryan_att,
17:55:10)
- Jamil: we can target a spec for license
management (bryan_att,
17:55:26)
- Bryan: we need to document the features at the
least (example will be put on the wiki) (bryan_att,
17:56:46)
- Ofer: may have issues with availability to move
this forward (bryan_att,
17:57:28)
- Jack: we need a team to work with Ofer to move
this forward (bryan_att,
17:57:47)
- Bryan: we need the projects and PTLs to
define/document what they intend for new features, so the community
committee can review those plans and comment (bryan_att,
18:00:00)
- Jack: feel free to reach out to us to help
build that architecture team to work on the features and related
product goals (bryan_att,
18:00:46)
- Jack: recommend Ofer kick off a document about
the process of documenting requirements (bryan_att,
18:01:50)
Meeting ended at 18:01:53 UTC
(full logs).
Action items
- (none)
People present (lines said)
- bryan_att (59)
- collabot_ (4)
- aimeeu (0)
Generated by MeetBot 0.1.4.