Skip to content

Optimize DB Step by Step

Every cube overnight.

Optimize DB doesn't benchmark. It sorts each cube's dimensions by leaf count, fewest first, and keeps the change only if the cube didn't use more memory (revert_on_regression, on by default). Use it to shrink a whole instance cheaply, then run optimize on the cubes that matter most.

Open the diagram full screen

In the UI

  1. Open Optimize DB, choose the instance and set the options: time limit, minimum cube size, cubes to exclude, and whether to disable active chores for the run.
  2. Click Build plan. Nothing on the server changes. Review the cubes it will reorder, their new orders, and the cubes it skips and why. Every cube is reordered by leaf count; for a cube whose densest dimension is not its largest, such as one with a fully populated months dimension, set its Method to Last slot only. That cube is measured, one reorder per dimension.
  3. Click Run plan and confirm. Progress shows cube by cube. Stop after the current reorder ends the run once the reorder in progress finishes.
  4. After a stop or the time limit, Previous runs offers Resume, which continues in the time that was left. If a run ended without re-enabling chores, the page offers Re-enable chores.
  5. See the outcome in the run's report: Report in Previous runs, or the run under Optimize DB (all cubes) on the Reports page. It lists what was reordered, reverted or failed, and the expected saving.

On the command line

optimuspy optimize-db instructions.json --dry-run     # build and print the plan, change nothing
optimuspy optimize-db instructions.json               # build and run
optimuspy optimize-db --plan <plan file>              # run a plan you reviewed
optimuspy optimize-db --resume <plan id>              # continue a stopped run
optimuspy optimize-db --restore-chores <plan id>      # re-enable chores a crashed run left off

samples/optimize_db.json shows the common options. To measure the last slot of some cubes, list them in last_slot_cubes. The plan, the run and the report are written to results/<instance>/, and the run summary ends with the report's path.

Note

After an Optimize DB run, restart TM1 before you benchmark cubes with optimize.

Every option is explained on Optimize DB Mode, and every control of the page on the Optimize DB Page.

Back to the start: Choose a Mode