Pipe

The pipe behavior of a given resource can be customized in case of more complex queueing operations are needed in addition to the simple FIFO discipline. With this purpose, multiple queueus can be added to the pipe and two decorators can be used to determine the enqueueing and dequeueing behavior of the pipe. The two decorators are @simpype.pipe.enqueue and @simpype.pipe.dequeue (see enqueue() and dequeue() for more details). There are two ways of customizing the behavior of a pipe:

• Work directly on the simulation scenario and perform an inline customization;
• Create a pipe model to be included in the simulation scenario.

Either approaches are valid, however inline customization is more suited for small customizations while pipe model is more suited for larger customizations and code re-usability (you can include the smae model multiple times in different simulations).

Inline customization

In this example, a priority queue with two service classes is implemented.

import simpype
import random

sim = simpype.Simulation(id = 'simple')
gen0 = sim.add_generator(id = 'gen0')
gen0.message.property['priority'] = {
    0: lambda: random.randint(0,1)
}
res0 = sim.add_resource(id = 'res0')
res0.pipe.add_queue('slow')
res0.pipe.add_queue('fast')
res0.random['service'] = {
    0: lambda: 2.0
}

@simpype.pipe.enqueue(res0.pipe)
def enqueue(self, message):
    if message.property['priority'] == 0:
        return self.queue['slow'].push(message)
    elif message.property['priority'] == 1:
        return self.queue['fast'].push(message)
    else:
        return message.drop('unsupported priority')

@simpype.pipe.dequeue(res0.pipe)
def dequeue(self):
    if len(self.queue['fast']) > 0:
        return self.queue['fast'].pop()
    elif len(self.queue['slow']) > 0:
        return self.queue['slow'].pop()
    else
        return None

sim.run(until = 10)

Custom model

Alternatively, a separate pipe model can be created to implement the same pipe behavior:

1. Edit mypipe.py with a text editor and create a pipe model in the following way:

import simpype

class MyPipe(simpype.Pipe):
    def __init__(self, sim, resource, id):
        super().__init__(sim, resource, id)
        self.add_queue(id = 'slow')
        self.add_queue(id = 'fast')

   @simpype.pipe.enqueue
   def enqueue(self, message):
       if message.property['priority'] == 0:
           return self.queue['slow'].push(message)
       elif message.property['priority'] == 1:
           return self.queue['fast'].push(message)
       else:
           return message.drop('unsupported priority')

   @simpype.pipe.dequeue
   def dequeue(self):
       if len(self.queue['fast']) > 0:
           return self.queue['fast'].pop()
       elif len(self.queue['slow']) > 0:
           return self.queue['slow'].pop()
       else
           return None

# Do NOT remove. This is required for SimPype to build your model.
pipe = lambda *args: MyPipe(*args)

2. Create your simulation scenario including the new model:

import simpype
import random

sim = simpype.Simulation(id = 'simple')
gen0 = sim.add_generator(id = 'gen0')
gen0.message.property['priority'] = {
    0: lambda: random.randint(0,1)
}
res0 = sim.add_resource(id = 'res0', pipe = 'mypipe')
res0.random['service'] = {
    0: lambda: 2.0
}

sim.run(until = 10)

3. Make sure that the file and directory structure is the following:

<working directory>
|-- simple.py
|-- mypipe.py

4. If you want to change the directory where SimPype looks for custom models, set the following variable in the simulation environment:

import simpype

sim = simpype.Simulation(id = 'simple')
sim.model.dir = '<your model dir>'

Please make sure you have reading permissions for <your model dir>. In this case, the file and directory structure would look like:

<working directory>
|-- simple.py

<your model dir>
|-- mypipe.py