24_nov_17 - sakura-team/sakura GitHub Wiki

Attendees

Mahfoud Boudis, Denis Bouhineau, Etienne Duble, Michael Ortega, Ritesh Shah

Duration

1h30

Addressed points

  • Discussion around IKATS and Sakura:

    • Sakura is more 'generic', i.e. accept more different data and analysis
    • Sakura could connect to IKATS for managing temporal data efficiently, and make joints between several kinds of data
    • History: we tried to bring together Sakura and IKATS, but we didn't succeed:
      • spatial data was not in the roadmap of IKATS
      • IKATS needs high-end hardware, and it was not possible to split elements
      • differences is the initial philosophy (at least at that time). E.g. IKATS was using push mechanism (Sakura is using pull), which we think is not scalable

  • Demo:

    • One can create a database
    • One can create a dataset/table (empty one for now, with column name, type and tags only)

  • Discussions in progress:

    • Do we keep Datas with a 's' ?

  • Meeting with Éric: we are not sure in which level of details Éric will question us, thus if we should prepare demo(s), slide(s), etc. Anyway:

    • We plan to show a demo about dataset management + (shortly) the link with workflows (e.g. twitterdb data can be the data source of a workflow of operators)
    • We will probably show the architecture diagram. In this diagram, we will replace the "Gricad platform" box with an "Ikats platform" box, since the later project seems more mature.
    • We will prepare a short set of slides showing what makes Sakura's data processing "special", if related discussion occur.

  • Todo :

    • Datas tab:

      • In ‘new Data’:
        • rename DBMS to Datastore
        • rename DBMS labels: ex: postgresql (driver label) service on idrap.imag.fr (host)
        • pass 'public' variable into the kwargs of the ws_request
        • in the datastores list: add ‘more’ for showing datastores that the user cannot access (offline or not allowed for this user)
        • display the database main page as soon as it is created
        • always display databases in the same order

    • Dataset module:

      • ‘New Dataset’:
        • description bug: put dataset instead of table
        • put the date of the day by default
        • refresh datasets list after the creation

    • General:

      • Only one 'sign in' should appear