Location

BrowserRoom SystemPhone Dial-in

https://bluejeans.com/293724745/

  1. Dial: 199.48.152.152 or bjn.vc
  2. Enter Meeting ID: 293724745 -or- use the pairing code

Dial-in numbers:

  • +1 408 740 7256
  • +1 888 240 2560 (US Toll Free)
  • +1 408 317 9253 (Alternate Number)

Meeting ID: 293724745

(back to the list of all DMLT meeting minutes)

Time

10:15am PT

Attendees

Regrets

DMLT Meeting Goals

Discussion Items

Guidelines: Propose items in line with the meeting rationale (see above); include as much information in the agenda notes; summarize, don't go into details while on the call; maximize time available for discussion. 

ItemWho

Notes

Project updates
  • A reminder that Leanne Guy is invited to the management meeting on Monday morning.
    • (But is not required to be there).
Test specs due in the next 45 days

  • One test spec overdue .
  • Unknown User (mbutler) Follow up on test plan for LDM-503-08b; make sure .  
  • Wil O'Mullane Clarify whether Jira tickets refer to test specs or test plans.  
Hardware BudgetWil O'Mullane
  • Need to take care to buy what we need (barely) so we have plenty for later.
  • Do we need all the K8 Nodes?
    • The intention was that the would be used for running Dask.
    • Supporting e.g. 40 stack club folks.
    • Can be scaled back to support fewer users doing fewer things.
  • Can we test Qserv at scale on Google?
    • In the short term, yes.
    • Longer term, there is risk exposure from not running on the system that would actually be deployed in operations.
  • Leanne Guy asks: where should enthusiastic science collaboration users be given access?
    • Not clear how this fits into the hardware budget.
Meetings next weekJohn Swinbank
Overdue RFCs
Change Control Board report
  • Change requests of interest to DM
  • HiPS image generation was approved, but then stumbled over costing; should now be resolved.
  • Kian-Tat Lim aiming to get header service LSE-400 this week.
    • Will include explicit header appendix (hooray!).
Open todo items


Report from Science Platform workshop
  • Walked through the LSP FDR charge, leading to submission of RFC;
  • Walked through User DB Workspace issues; Leanne asked to provide formal set of use cases; must be digested into design in time for the review;
  • Reviewed authorization architecture alternatives; rejected "pure capabilities-based" architecture as poorly suited to needs, at least with our resources; KT to write up memo with BvK on the pure-identity and mixed models discussed, with pure-identity, group-membership-based auth'z'n as an initial phase;
  • Reviewed new plan for internal and external use of IVOA standards (see DMTN-090); Gregory will RFC as a change to LSP Design, LDM-542;
  • Adopted CADC TAP service Java code base as the core of our planned TAP service (following Christine's nice demo); Fritz to document this and describe the development required to fully adapt/extend it to our design (edited)
  • Examined demo of Dask integrated with Jellybean; discussed Dask, Spark, and other "next-to-DB" issues
  • Looked at demos of closer Firefly-JupyterLab integration and of Firefly LSST detection footprint browsing

(The above are lifted from Gregory Dubois-Felsmann on Slack)

  • Progress working through A&A issues.
    • There will be a memo forthcoming from Kian-Tat Lim et al on this.
    • This will be at a level which enables detailed costing.
  • There are action items on the meeting page; immediate aim should be to address those.
DMLT travel & availability this week
  • John Swinbank in the UK all week; available but seven hours shifted.
  • Wil O'Mullane in Chile traveling back Wednesday night (so not really available Wednesday Tucson working time).
  • Margaret Gelman at LBNL for DESI operations review
No Monday telecon next week due to the DMLT face-to-face meeting.

Past action items

Open DMLT actions