Skip to content

aybabtme/desim

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

desim

Discrete Event Simulations

Focus on your model, desim handles the rest.

about

desim is a framework that allows one to quickly produce programs that implement a simulation. desim takes care of scheduling events in a simulation and gathering results, and lets you focus on the model you want to simulate.

desim is a simple package for one off simulations, but long term is meant to become a development environment platform for distributed simulations.

example

If you model a clock like this:

func clock(iter int, dur time.Duration) desim.Action {
	pdur := gen.StaticDuration(dur)
	return func(env desim.Env) bool {
		iter--
		env.Log().Event("woke up, about to sleep")

		if env.Sleep(pdur) {
			env.Log().Event("couldn't sleep")
			return false
		}
		return iter > 0
	}
}

And would like to play a simulation with 2 clocks, a fast and a slow one:

slowActor := desim.MakeActor("slow", clock(6, 1*time.Second))
fastActor := desim.MakeActor("fast", clock(6, 500*time.Millisecond))

You can! Just create an event scheduler, some parameters for time, and run your actors to see how they would behave:

var (
	r     = rand.New(rand.NewSource(42))
	start = time.Unix(0, 0).UTC()
	end   = start.Add(10 * time.Second)
)

sim := desim.New(
	desim.NewLocalScheduler,
	r,
	gen.StaticTime(start),
	gen.StaticTime(end),
)

history := sim.Run(
	[]*desim.Actor{
		slowActor,
		fastActor,
	},
	desim.LogJSON(ioutil.Discard),
)

Where history represents what events were scheduled, and desim.LogJSON records what the actors have emitted during that time (in our case, that goes to /dev/null).

performance

Simulations happen in simulated times: time goes as fast as possible. The speed at which the simulation runs grows linearly with the number of events happening over time.

For instance, if we simulate 1 hour worth of N actors performing events every 500ms, we see that the time it takes to simulate this hour is linear with the number of actors involved (and thus, with the frequency of events).

Time to simulate 1h worth of events

Similary, if we have a single actor simulation, but with the actor performing events at a fixed frequency, we see that this linear relation holds:

Time to simulate 1h worth of events

license

MIT license