• Investigating job slowness compared to other sites in LSST (RAL 3:10:12,  IN2P3 1:17:53, LANCS 1:39:05)
  
  • IO from echo?
    • Current thinking from LSST is that reading files from echo takes twice as long as other sites
      • 0.6s per file at RAL, over 0.3s at others. Not much but over 20,000 files adds up! (And this is a small test data set not a real one)
        • LANCS mentioned today in the storage meeting they are creating a pool just for LSST with 4+2 to try and improve performance for LSST
  • CPU / memory bound?
• Request to all the sites if there is a shared space at their site to create a sqLite database within a DAG (job of jobs) to track progress within the DAG
  • Informed requester of infrastructure at RAL and they seem happy that the local scrath for jobs and ECHO would be fine
• Will begin movement of ComCam data to RAL this week