question

Pass different stream values for each distribution in your MTBF and MTTR tables.

You are using the same random number stream for all of your distributions. By doing this, if the timing of events changes the order that various stochastic distributions are called, then the values returned by subsequent calls will be different.

You can avoid this type of unintended relation between seemingly unrelated elements of your system by using different values for the stream parameter on each random distribution you use. That is what the stream parameter is for.

Here's a simple example to illustrate what is happening:

Imagine that random stream 0 returns the following random values: 3, 2, 7, 5, 4, 5, 6, 3, 9, 8, 1. Then imagine that these events happen in a particular order and each sample random numbers using stream 0.

event : random number sampled

• A source interarrival time : 3
• B processor process time : 2
• C processor process time : 7
• D mtbf : 5
• E source interarrival time : 4
• F processor process time : 5
• G mttr : 6
• H processor process time : 3

Then imagine that you changed something about your model (such as the queue's size) so that those events changed order:

• A source interarrival time : 3
• B processor process time : 2
• D mtbf : 7
• C processor process time : 5
• E source interarrival time : 4
• G mttr : 5
• F processor process time : 6
• H processor process time : 3

Now the values of the mtbf and mttr stochastic functions are based on different random numbers, even though you only changed a seemingly unrelated element in the system.

By specifying different stream values for each distribution, the order of events in the other parts of the model will not change what random values are being returned for that distribution.

For more information, see Averill M. Law's book "Simulation Modeling and Analysis" 4th Edition Chapter 11 "Variance-Reduction Techniques" pages 584-586.