• Training server actually runs in BangDB server framework
  • Train stage will do training locally
  • The servers could be behind LB (Logical)
  • Train stage gets training data from BRS
  • Training server keeps the training related metadata at BRS as well (BangDB)
    • Param & Model info, Size of model
    • Algo used, Time took to create model
    • Last updated timestamp, versions
    • Model location, Model
    • Training & other files etc.
  • Train stage may also be part of server where store stage is RS + Data source
  • Native integration ensures data comes from DB
Deep Learning Diagram