14:02:23 #startmeeting Acumos TSC Architecture Committee 14:02:23 Meeting started Thu Jul 26 14:02:23 2018 UTC. The chair is bryan_att. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:02:23 Useful Commands: #action #agreed #help #info #idea #link #topic. 14:02:23 The meeting name has been set to 'acumos_tsc_architecture_committee' 14:08:14 #topic Pantelis: Community approval of Data Pipeline and Model Training Pipeline features for Release A & B 14:17:08 #info per the Jira items ACUMOS-1367, 1420, 1466, 1418 14:17:38 #info and description at https://wiki.acumos.org/display/TRAIN 14:18:28 #info and specifically https://wiki.acumos.org/display/TRAIN/E2E+Project+Requirements 14:22:07 #info For data pipeline, https://wiki.acumos.org/display/TRAIN/DP+Requirements 14:23:24 #info Bryan: additional explanation on what "kubernetes native APIs" and the role of the Acumos platform in supporting "diverse environments" for aspects of the data pipeline, would be helpful for understanding and agreement with these requirements 14:24:41 #info Bryan: an architectural concept would also be very helpful to clarify the role of the Acumos platform and the interfaces it needs to support to these diverse environments, and how those interfaces might depend on "kubernetes native APIs" is that's part of the intent of that requirement 14:29:18 #info Anwar: how are ETL components linked to training? 14:30:07 #info Anwar: how will this fit into leveraging CMLP code? 14:31:00 #info Adi: for data pipline, we have some components that can be leveraged. We have kafka for example, but need to understand the overall experience expected, to be sure we can leverage the assets we have implemented. 14:31:12 #info Anwar: any of those reqs planned for Nov release? 14:31:26 #info Adi: already have data set mgmt capabilities 14:31:47 #info Anwar: for Acumos Nov release? 14:32:05 #info Adi: that's unclear yet, need deeper dive on that 14:32:26 #info Anwar: Pantelis, recommend you participate in that discussion 14:33:05 #info Pantelis: it's important to ensure the DP arch moving forward can be supported in all the envs using a cloud-native approach 14:33:29 #info Anwar: is CMLP implementation k8s-based 14:33:46 #info Adi: yes 14:34:04 #info Bryan: does this mean cloud-native, or just that it runs under k8s 14:34:11 #info Adi: just runs under k8s 14:36:17 #info Bryan: e.g. functional resiliency, microservice-oriented means specific things about the ephermerality of services - and will cause us to need to restructure/refactor platform components probably 14:37:01 #info Pantelis: moving on to https://wiki.acumos.org/display/TRAIN/MTP+Requirements 14:47:12 #info Anwar: for the Community Committee, has there been a review and agreement with these requirements? 14:47:29 #info Ofer: not specifically 14:48:10 #info Anwar: there may be some things e.g. data set management that the CMLP team can commit for Athena - but we will check 14:48:28 #info Manoop: will all of this be CMLP seed code? 14:48:51 #info Anwar: only the data set management, maybe other things, but this is much larger 14:49:26 #info Manoop: assume most of this is for future releases; hat team will develop this? 14:50:01 #info Anwar: need to hear from Nat or otherwise know what related to this will be delivered in Athena, high-level epics at least 14:50:40 #info Nat: we have requested the input from PTLs and have some, but not all 14:50:53 #topic Portal API review 14:51:37 #info Mukesh: shares https://wiki.acumos.org/display/PUX/Athena+M3+Portal+Deliverables+for+API+Freeze+Milestone+Checklist 14:54:34 #info Bryan: the API spec needs to define the request/response objects 15:07:00 #endmeeting