Skip to content

Reading the Result

What the report tells you, and where the cube ends up.

  • Best Overall is the recommended order. OptimusPy checks the orders in the order they were tested, starting with the original. The first one within 1% of the best on memory, and on query and TI time where you gave views or processes, wins. The 1% is of the gap between the best and the worst order tested. If none qualifies, it tries again at 2.5%, then 5%. If there is still none, the original order is put back and the log asks you to pick one from the results. In a run that measures RAM only, there's nothing to balance, so Best Overall is simply the lowest-RAM order, unless the original is within the 1%. Either way, the original order wins whenever nothing is clearly better.

    Report podium with the Best Overall, #1 Fastest Query and #1 Lowest RAM cards

  • The chart plots every tested order: memory across, query time relative to the original up. Lower and further left is better. With processes and no views, the vertical axis is TI time relative to the original instead. With neither, it shows the memory saving, so higher is better.

    Scatter chart of every tested order, RAM against query time relative to the original, with the original order and the result marked

  • "Not measurable". If no tested order changed memory at all, the log warns that the cube wasn't measurable: the orders can't be told apart on memory, so pick one yourself from the timings.

  • Where the cube ends up. With auto-apply ("update": true; the UI writes "auto_apply": true, which means the same) the best order is applied. Otherwise the original is put back and you apply the winner yourself, with set or Sync Order. See Taking an Order to Production.

Every part of the report is described on the Reports Page, and the selection rule on How the best order is chosen.

Next: Taking an Order to Production →