Child pages
  • 1st April 2015
Skip to end of metadata
Go to start of metadata

Attending: Scott Daniel, Peter Yoachim, Michael Reuter, Darko Jevremovic, Cathy Petry, Kem Cook, Rahul Biswas, Simon Krughoff, Veljko Vujcic, Chris Walter, Brittan Fiore-Silvfast, Lynne Jones

This weeks phonecon will be run by Scott and will cover

  - demo of DDS (Michael)

        - how will it scale to running at 1000x times realtime

  • DDS = how components of the control system will communicate with each other over ethernet.  Uses "publish/subscribe" formalism.
  • Only runs on CentOS 7 right now
  • Has very specific firewall requirements (not a problem if we just run it on Docker containers; but if people want to eups install it locally, they will have to change their firewall configuration)
  • There seems to be a limitation on rate.  Can only push messages out every 30 milliseconds (or slower) if you want the subscriber to receive all of the messages
    • Kem thinks this could be a problem for passing data around inside of the scheduler.
    • Michael will talk to David Mills about this.  Maybe there is an easy way to fix it.
    • Kem thinks Srini has a fast control loop for guiding and seeing; maybe we can learn from his experience.
  • Need another CentOS 7 machine to test machine-to-machine communication
    • David Mills is working on making DDS compatible with other OSs

  - discussion of new baseline cadence run (Kem, Lynne, Zeljko)–brief description of 3.61 & enigma_1189

        - can we sign off on the new baseline for Zeljko to present to the CCB

  • new reference run more repeatable than old reference run
  • proposals refactored to better enable deep drilling fields as compared to old version
  • seem to have eliminated end-of-survey "panic" (rapidly changing filters and position during last year)
  • Lynne thinks enigma_1189 is better than OpSim3_61: it has a more realistic down time distribution
  • From Lynne:
    Tusken has some MAF analysis results from enigma_1189 :
    the "scheduler validation" metrics (on opsim field pointings, not healpix grid .. looks at distributions etc relevant for scheduler)
    the "science performance" metrics (on healpix grids, looks at more specific science goals .. just things described in SRD so far)

Other business

OpSim stand-ups will be 10am Pacific Time on Mondays and Thursdays


  • No labels