Leanne Guy

• Leanne Guysend save-the-date notes out 10 May 2019 

Colin Slater

Melissa Graham  is leading the SOC and Colin Slater  is on the POC. Some possible ideas (MLG seeded draft session proposals below, please update them): 

1. User-Generated Data Products
   • We concluded that there is not enough concrete implementation of a system for UG data products yet to warrant a session on this topic. 
   • We have presented the requirements many times but the interfaces are not ready
   • WebDav is maturing but nothing concrete yet
   • Proposal & implementation  from CADC on providing via TAP service
   • Does not cover all use cases but in moving in the right direction
   • No DB backend yet
   • Basically nothing demo-able
   • Conclusion: too early to  show any sort of demo on this yet 
     
     
2. Photometric redshifts (MLG will liaise with DESC PZ WG)
   • Conclusion: Submit a proposal for a session. 
   • Melissa Graham will lead this from the DM side, Jim Bosch Colin Slater Leanne Guy Zeljko Ivezic should also plan to attend. Robert Blum will be invited as well. 
     • DM will communicate the current baseline and issues as we see it such as: 
       • Do we just take a published algorithm or someone's table and publish it?
       • Do we train models or do we take a trained model from the community?
       • How will photo-Zs be validated in commissioning?
3. Run a session during the time-domain day-of-focus on working with time series in the LSP using precursor data 
   • We have stripe83 and WISE catalogs but no time series model to show  
   • IVOA have done a lot on TS data models – will be a focus at the Paris meeting next week
   • What do users expect in terms of a time-series data model and associated attributes on time series?
   • Feedback from the LSP review is that they want not just aggregated information and time series, but also ready access to postage stamps for the time series. They did not like the idea of batch service for retrieving a bunch of objects. 
   • Conclusion:  we are not ready to show any sort of demo on this yet but will participate in a requirements gathering session around working with time series as part of the time-domain data products session by Michael Wood-Vasey
   • Gregory Dubois-Felsmann Melissa Graham Eric Bellm Leanne Guy at least should be present. All welcome.
     
     
4. Team up with stack-club folk to run a tutorial session on doing LSST science with the LSP 
   • Gregory Dubois-Felsmann is worried about creating expectations that we cannot meet. 
   • The stack-club model is good in that people who use the platform contribute something  (analysis notebooks) - it is not just a free-for-all
   • Phil  is open to extending this to include LSP usage and not just the stack. Perhaps as one team or 2 separate (LSP/Stack)
   • Conclusion is that we will wait until we have HSC and Gaia data and re-consider this again next year
     
     
5. A deep-dive into the LSST algorithms would be great but that is what the workshop later this year is for and I don’t want to do the same thing twice. We should have a good idea of the status of 
   • No disagreement

• Leanne Guyto submit proposal for session Photometric Redshifts in the Data Release Data Products (Melissa Grahamis on holiday) 13 May 2019 
• Leanne Guy make sure Bob Blum is apprised of plans for a photo-z session at the 2019 PCW  17 May 2019 

Robert Gruendl

Summary of what happened 

• Transferred raft scale data from base to NCSA each morning, ingested into a Gen2 butler and ran through processCCD (ISR) 
• rsync with a file sent every 20sec. total throughput 2MB/s
• More like a normal observing run and not commissioning - things will change more frequently in commissioning. 
• latency for base → NCSA transfers:  1 sec base 

• Colin Slater added that many lessons were learnt on the science side

  • It is important to have good QA and visualization to see how the night proceeds rather than a big results file to go away and look at. This should be raised to the successor of the QAWG

  • There were something like a few dozen failures in a thousand (failure == fail to process an image to produce photometry)

  • Lauren understood the causes  immediately. Bugs were already known and Jira tickets exist already

  • Robert Lupton points out that the correct process is to process all the images to completion w/o fail and apply science criteria at the end

Following related discussion on data transfers for AuxTel data, the topic of latency and access to AuxTel data will be raised at the upcoming DMLT-F2F

• John SwinbankAdd agenda item for AuxTel data transfers, ingestion and access via LSP to agenda 17 May 2019 

Colin Slater

Colin Slater has made a start on the  algorithm classification

Some updates made during the meeting - SST folk are requested to add their inputs before 24 May 2019 . The goal is to identify which of these the SST can contribute to,  assessing options to identify a course of action. 

Colin Slater and Leanne Guy will use this to build an agenda for the F2F meeting.

• Colin SlaterAdd  connection to LDM-151  to the algorithms classification page  17 May 2019 

Leanne Guy