Retrieving Simulation Configuration Data From Audit Trail Databases

Data about the behavior of the system to be simulated such as

• duration distribution of activities and/or applications,
• availability of participants and/or applications,
• arrival rates of process instances or
• probabilities of transition traversals

can be retrieved from Stardust audit trail databases. This way, you may predict the future behavior of your system based on real-world data.

Retrieval Configuration

Select the Audit Trail Retrieval tab of the Simulation Runtime Configuration.

The following information needs to be specified for the database connection:

• Database Type for the type of the database used,
• Driver Class Name for the class name of the JDBC driver class for this database,
• Database URL for the URL of the database,
• Account for the database account and
• Password for the password for this database account.

Evaluate Audit Trail Data before Simulation Run to specify whether the audit trail driven parameters should be evaluated up front to any simulation run, whereby Ask for Confirmation opens up a confirmation dialog up front to the simulation run which allows to specify whether parameters should be retrieved from audit trail or not.

Start of Evaluation Interval and End of Evaluation Interval to specify the retrieval interval for audit trail data. The retrieval interval is independent of the simulation configuration interval. It may end before the simulation configuration interval or may even overlap with it.

Interpolation Unit for Daily Availability describes the quantisation of the daily availability distribution. It may vary from 15 minutes up to 8 hours (see below for details).

Interpolation Unit for Available Resources describes the quantisation of the long-term distribution (see below for details).

Default Interpolation Mode used to obtain multi-day availability curve (see below for details):

• Maximum Value should be specified if all resources having actually been available should be considered in the simulation configuration.
• Average Value should be used if all resource having actually been needed should be considered in the simulation configuration.

Default Approximation Method approximation algorithm (see below for details) with the options

• Translation of the Starting Point
  to the beginning of the simulation interval. This option is useful when you intend to simulate a specific event in the context of a real-world availability distribution.
• Repetition
  of the curve until the end of the simulation interval. This approach is useful when the retrieval interval is a comprehensive, self-contained cycle, e.g. an entire year.
• Linear Approximation
  via least square approximation assuming linear growth of availabilities.

Reset Availability Calendar option to decide whether the business calendar is deleted during retrieval, so all days are set to working days, or if all manually set holidays are kept. The default value is true.

Duration Distribution Retrieval

For duration distribution retrieval, the duration of an activity is determined considering all instances of this activity. For applications all instances of activities invoking this application are considered.

Normal Distribution

For the Normal distribution, average and standard deviation are retrieved.

Poisson Distribution

For the Poisson distribution, the average will be retrieved and rounded to the next integer value.

Uniform Distribution

For the Uniform distribution. the minimum and maximum values are determined.

Custom Distribution

For custom distributions, the complete histogram of the custom distribution is retrieved. Hereby, an intelligent algorithm is used to determine the histogram interval based on the variation of values.

Resource Availability Retrieval

As a prerequisite to all further computations, the total workload curve for the resource is computed, e.g. for 3 days as sketched in the following diagram:

Daily Resource Availability

To determine the daily resource availability,

1. the curve is split in single days
2. the daily curves are normalized assigning 100% to the maximum value
3. the daily curves will be overlayed,
4. the points of the curve will be grouped in appropriate intervals (e.g. hours),
5. the average or maximum of an interval will be used as the resulting value per interval

The following example shows how the daily distribution is determined from the 3-day distribution:

Click to enlarge.

Available Number of Resources

1. The original total workload curve will be separated in intervals (e.g. days or months) and
2. a single value per interval is calculated according to the interpolation mode.

The following diagram shows an example with an interval length of 1 day:

Click to enlarge.

Approximation into the Simulation Configuration Interval

According to the approximation method all curves are transformed to their values in the simulation configuration interval.