I am new to JMT and queueing theory in general so apologies if this is a stupid question.
I am trying to model a system, part of which needs the customers to get cycled around a loop with a definable number of cycles.
I can't see how to do this with JMT as none of the routing strategies seem appropriate.
I am actually trying to model a more "real-world" processes rather than anything computational.Perhaps there are other tools out there that would be more appropriate for my purpose?
Best regards
Gareth
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello Gareth,
you can use the "probability" routing strategy.
For example if customers need to loop 3 times before exiting the system you can place 0.75 probability of looping and 0.25 probability of being directed out from the loop.
Considering just a single request this is not guaranteeing 3 exact cycles but, statistically speaking and given the large number of analyzed samples in the simulator, it should model well your environment.
Best Regards,
Marco
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Many thanks for the reply. I did think about using probability based routing. It is not how the real-world system will work but might be good enough for modelling.
However the loop contains a queueing network parts of which would be sensitive to becoming "jammed". Hence I worry it would be sensitive to such randomness in the system.
Best regards
Gareth
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Gareth,
in alternative to the suggestion of Marco you may also use the following technique that consider the Class Switch (CS) feature of JSIMgraph. For example, if you want to have 4 cycles, you may define 5 classes of requests: A,B,C,D,E. The 5 classes differentiate only for their routing probabilities and not for the other properties. The fifth class is dummy. Then, locate a CS station at the end of the loop. If request A is arriving, then with prob.1 will become of class B, if B is arriving then will become of class C, etc. If class D is arriving then it will become of class E. The routing probabilities of the CS station will be: Class A, B, C, D to the begin of the loop with prob.1, Class E to the exit of the loop with prob.1. You will sure that a request will loop for 4 times. If the model is open, set the interarrival times of the classes B,C,D,E very very large, while that of class A the one you have in the experiment.
Well, hope you are able to obtain good results for your problem!
Best regards
Giuseppe
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Many thanks for the reply.
In the real world system the number of cycles can reach into 1000s so generating so many classes would be impractical!
However, after giving it some more thought I can simplify the model without changing the results (I think).
Many thanks for the help and excellent tools!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Gareth,
I undestand. We should think on a scheduling policy that accept a fixed number of exit per path. However, there are several decisions that should be made with this policy. What happens when the max number of cycles is reached and there are several other routing paths? And, how to control if internally in the loop there are other stations with fixed exit policy with parameters that are not possible to satisfy due to the other controls on the number cycles?
I mean, consistency checks are impossible unless we keep under control the complete loop, and this is really a huge problem.
Anyway, thanks for your requests and for using JMT!
best regards
giuseppe
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi
I am new to JMT and queueing theory in general so apologies if this is a stupid question.
I am trying to model a system, part of which needs the customers to get cycled around a loop with a definable number of cycles.
I can't see how to do this with JMT as none of the routing strategies seem appropriate.
I am actually trying to model a more "real-world" processes rather than anything computational.Perhaps there are other tools out there that would be more appropriate for my purpose?
Best regards
Gareth
Hello Gareth,
you can use the "probability" routing strategy.
For example if customers need to loop 3 times before exiting the system you can place 0.75 probability of looping and 0.25 probability of being directed out from the loop.
Considering just a single request this is not guaranteeing 3 exact cycles but, statistically speaking and given the large number of analyzed samples in the simulator, it should model well your environment.
Best Regards,
Marco
Dear Marco
Many thanks for the reply. I did think about using probability based routing. It is not how the real-world system will work but might be good enough for modelling.
However the loop contains a queueing network parts of which would be sensitive to becoming "jammed". Hence I worry it would be sensitive to such randomness in the system.
Best regards
Gareth
Hi Gareth,
in alternative to the suggestion of Marco you may also use the following technique that consider the Class Switch (CS) feature of JSIMgraph. For example, if you want to have 4 cycles, you may define 5 classes of requests: A,B,C,D,E. The 5 classes differentiate only for their routing probabilities and not for the other properties. The fifth class is dummy. Then, locate a CS station at the end of the loop. If request A is arriving, then with prob.1 will become of class B, if B is arriving then will become of class C, etc. If class D is arriving then it will become of class E. The routing probabilities of the CS station will be: Class A, B, C, D to the begin of the loop with prob.1, Class E to the exit of the loop with prob.1. You will sure that a request will loop for 4 times. If the model is open, set the interarrival times of the classes B,C,D,E very very large, while that of class A the one you have in the experiment.
Well, hope you are able to obtain good results for your problem!
Best regards
Giuseppe
Hi
Many thanks for the reply.
In the real world system the number of cycles can reach into 1000s so generating so many classes would be impractical!
However, after giving it some more thought I can simplify the model without changing the results (I think).
Many thanks for the help and excellent tools!
Hi Gareth,
I undestand. We should think on a scheduling policy that accept a fixed number of exit per path. However, there are several decisions that should be made with this policy. What happens when the max number of cycles is reached and there are several other routing paths? And, how to control if internally in the loop there are other stations with fixed exit policy with parameters that are not possible to satisfy due to the other controls on the number cycles?
I mean, consistency checks are impossible unless we keep under control the complete loop, and this is really a huge problem.
Anyway, thanks for your requests and for using JMT!
best regards
giuseppe