Dear community members,
I crated a model with 7 sources (model 01). Each source has its own continuous empirical distribution and seed number (seed 1 for the first source, seed 2 for the second …). The source produces a customer and each customer sets the wanted product quantity in a user defined attribute called “random”. The wanted quantity is recorded in a table called “wantstable_simplified”.
I created a model copy (model VS2) and the only different thing is the parallelproc cycle time. In this model (model VS2) the customer coming from the first source (sourcecheeseburger) generate the same numbers in the wantstable as in the first model, while the customers from the second source (sourcePromo) generate the same number until customer 5, then the numbers change. I noticed that this happens when more than one customers from sourcePromo are in the model.
Since all the sources have their own seed number, all the numbers in the wantstable should be equal. Why it is not the case?
I attach the models for reference and if you are going to check them, suggest to see what happens in model VS2 at the 6th customer from sourcePromo.
Solved! Go to Solution.
you set the wanted quantity in the exit control of your buffer after the source. The interval and the calculation of the wanted quantity both use random stream 2.
Now it depends on the number of parts in the buffer which distribution gets which random numbers.
Just an additional remark how to solve this:
Instead of using a formula with "z_cemp(2, emppromo)" for the interval at the source set the interval type to "cEmp" and specify the table emppromo.
And instead of calling "z_uniform(5, 1, 171)" in the method random create an additional user defined attribute at your MU of type "randtime". There you can then define your uniform distribution.
This way you get rid of your explicit management of random streams. The system now guarentees that you have independant random numbers.