Announcement Announcement Module
No announcement yet.
SDN performance differences between advanced and simple mapping Page Title Module
Move Remove Collapse
Conversation Detail Module
  • Filter
  • Time
  • Show
Clear All
new posts

  • SDN performance differences between advanced and simple mapping


    I'm struggling with SDN 2.1.0.M1, advanced mapping and STS 2.9.2. for a pair of days. When it works it's wonderful, but when something goes wrong I hit my head against the wall. There's a lot of aspect magic for me and I not comfortable with it, I have to keep ability to understand why things go wrong.

    I have to make a choice: keep going on the rough aspectJ road, or fall back to simple mapping.

    My criteria is performance. Michael write that aspectJ is the good choice to have performance.
    Is it possible to have an idea of difference between simple and advanced mapping from this point of view ?


    Marc DeXeT

  • #2

    Are you using embedded or server?

    In the server case, the first thing is always to minimize network traffic - advanced mode is better suited to the embedded use case.

    As for general rules: it depends But worth doing a simple model of your domain and then testing, before you lock yourself into one or the other.



    • #3
      Thank you.
      I'm planning to work with an embedded server, and I have a lot of write/read to do. I'm trying to create a referential for 120 000 items and I'm used to work following a idempotent pattern:
      1) get data from business referential for this item (datas are pushed or pulled following business sources)
      2) build in memory the representation of item (future state) from data
      3) read actual representation from graph (previous state)
      4) compare both of them
      5) apply differences if any
      6) save change log

      I have to compare node and related node to detect any difference so performance is not a minor aspect for me.

      But I'm going as you say: try and see



      • #4

        With embedded, you then have the choice of programming model - AspectJ can be difficult in some IDEs. As for performance, it shouldn't matter - 120000 records should fit in memory easily.