Getting started with ARIS simulator

Carsten,

I work for Lanner Group who provide the simulation component embedded within ARIS Business Simulator. I have posted a small number of articles relating to simulation which may be of interest. I would advise that you first experiment with a small EPC model, instantiating tokens, allocating process times and resources to constrain the model and check out the simulation. For any particular process question where ytou look to use simulation, good advice is to define the model at the right level of detail....dont go too complex fopr the problem being studied.

When you mention that you want to assign variables, what variables do you mean, what do you need to represent?

Geoff

Carsten,

Apart from the information in the Help, which is quite comprehensive in my opinion, another good source are the two examples in the UMG Demo database. One of them is for EPC simulation and it really uses a big part of the method. There are 13 models in the group /United Motors Group/2. Processes/Process Architecture/Core Processes/Procurement/Vehicle management/Receipt of goods/replacement parts for vehicles. You have most of the instantiation, recourse assignments and simulation control options shown in the example options.

The other is for BPMN 1.2 simulation. It illustrated with the process "Produce car door".

Both examples are made to focus on identification of bottle neck (seen by the high dynamic wait time at one of the activities) which is the cause if increasing throughput time, bad utilisation of resources and other problems (for example one of the departments working after-hours). But they can be used to learn about other applications of dynamic process simulation as well.

Hope that helps a bit though not as much as availability of free simulation tutorials would have helped.

Dear Mr. Geoff,

Greetings from my side. I have read few of your articles related to Simulation in ARIS. We have currently acquired ARIS Simulation and have managed to Simulate few of our processes. Currently, we have managed to showcase the Time consumed bewteen As-Is process and a suggested To-Be process.

Now, we have plans to move towards Cost Evaluation. Our focus is to define the cost associated with each Person/Role in the Organizational Chart. These Roles are already attached to processes within EPC models. We have already provided the Time it takes to perform each function and now want the system to calculate the cost of each function based on the cost of the attached Role.

However, I have not been able to find a function to support our requirement. I have only been able to provide Cost at each function and the system just calculates to total Cost for the whole EPC.

I hope you can provide guidance in this regard.

Thanks and Regards,

Anosh

Anosh

Human Resources (detailed) statistics should provide costs by Task. In my simulation work often I do not gather human resource costs within the simulation by task. Rather I use the degree of utilisation or number of times a task has been activated as a way of knowing the time spent, then use that as a driver for cost calculations externally. There are always different ways of doing the analysis!

If you have allocated costs at a detailed level in the Org Chart, and depending on how you allocate resources to tasks - on ocasions you might get expensive resources allocated to a task or cheaper ones in a fairly arbitary way....this is one reason I sometimes do costing outside using a system of 'standard costs'.

My comapny Lanner also do simulation of facilities...ports, LNG  terminals, pipelines etc using our WITNESS simulation, http://tinyurl.com/3qsb7tq

  

http://tinyurl.com/3qsb7tq

This may be of interest to others in your company,

best regards

Geoff

Thank you very much, with this input and the help of the global support I think I am off to a good start.

thanks

carsten

I just wanted to provide an update on where we are with the simulation. So the situation was that we have a situation where experiencing too long wait times for processing jobs on  one of our supercomputers, and the common assumption was that it was running at max capacity. So we wanted to see what would happen if we added extra capacity to this supercomputer.

The first step was to create a model of the as-is where we used the statistical attributes of the actual jobs processed in March as input data. After setting the variables and running a simulation equivalent to 3 months I could not get the simulator to show a full utilization of resources. After trying a couple of different scenarios we met with the system admins and we found out that there was wasted resources due to the fact that jobs were processed on a first come first serve basis rather fitting smaller jobs into the production when processors became free. Additionally short run jobs of less than five minutes could end up waiting behind long running jobs giving the users a poor experience. So the simulator showed us that the common assumption that we were running at 100 % was not entirely true, and we learned that we were not giving the best possible user experience. Users with jobs that ran more than 1 week are more tolerant to a wait time than a customer that has a job that only runs for a few min. we also learned that we could process 50% of our jobs just using 2% of the processing capacity and that equates to a lot of happy customers at a low cost. After confirming these findings we set up a new concept which we simulate the new queuing and found a better balance.

While there are still many things that we can do to improve the utilization we learned that with simulation we can test common assumptions, and that we could try out different scenarios without too much effort. While we can still use the additional resources to serve our long running jobs we have already improved service to the majority of customers with an absolute minimal negative impact on a few users.

Getting started with the simulator was easier than expected and we had a very simple but reliable “as-is” model in 2 days and setting up the new (also very simplyfied) scenario took less than one day. So the customers saw the benefits within two weeks.