Parallel Running of Dimensions

Added in Q4 22

The processing of the dimension results is parallelized to improve performance. There are three main stages:

1. Caching the base and analysis selections (used in the profiling of all dimensions)

2. Evaluating the Point-in-Time expression (each person’s event timepoint to aggregate their behaviour)

3. Running the profiles for each dimension in batches of 10 (by default)

The above stages are run sequentially since step 1 is a prerequisite for step 2, and both are required by all the profiles run in step 3. The profile batches are, however, submitted in parallel with a default of three batches running at any one time. As each batch finishes another is submitted to the job queue.

These settings can be adjusted for an individual Modelling Environment by viewing the Properties window:

The defaults give the following behaviour:

 

For more on working with dimensions see:

• Modelling Environment: How do I create a Behaviour-based Model?

• Selecting and Tagging Dimensions

• Using Measures of Association to find Diverse Dimensions

• Invalid Dimensions