Skip to content

Latest commit

 

History

History
112 lines (92 loc) · 5.78 KB

README.md

File metadata and controls

112 lines (92 loc) · 5.78 KB

Navigation: DEDIS :: Cothority Template :: Simulation

Simulation

After you know that a new service and protocol work at all (unit testing and integration testing have passed) then you might want to know how that software will behave in larger networks of 10's, 100's or 1000's of nodes. This is the job of the simulation system.

References:

Simulations in onet are a powerful way of making sure your code is well behaving also in bigger settings, including on different servers, and of course to write simulations used in research papers to make pretty graphs. In order to write a simulation, you must make a struct that implements the onet.Simulation interface.

You will need to implement the Setup method to return the *onet.SimulationConfig instance and to create a roster and a tree. The Setup method is run at the beginning of the simulation, on your computer. It prepares all the necessary structures and can also copy needed files for the actual simulation run.

The Node method is run before the actual simulation is started and is called once for every node. The simulation framework makes sure that all nodes have finished their Node method before the Run is called.

The Run method is only called on the root node, which is the first node of the Roster.

Running your simulation

The simulation example is in the directory simulation. It demonstrates how to write simulation drivers and configurations for simulating the behaviour of the protocol alone (files protocol.go and protocol.toml) or for simulating a client talking to a service (files service.go and service.toml). Of course, in our case we know that this example service always starts one instance of the example protocol, so the TemplateService will be exercising the protocol as well.

You can build the simulator executable with go build. If you try to run it with no options (./simulator), it asks for a simulation to run. You must give it one or more toml files on the commandline.

The protocol.toml file has:

Simulation = "TemplateProtocol"
Servers = 8
Bf = 4
Rounds = 10
CloseWait = 6000

Depth
1
2

When you run ./simulation protocol.toml this is what you get:

$ ./simulation protocol.toml
1 : (                        simul.startBuild:  54) - Deploying to localhost
1 : (           platform.(*Localhost).Cleanup:  92) - Nothing to clean up
1 : (             platform.(*Localhost).Start: 161) - Starting 8 applications of simulation TemplateProtocol
1 : (                       platform.Simulate: 117) - Started counting children with timeout of 1000
1 : (                       platform.Simulate: 121) - Found all 5 children
1 : (                       platform.Simulate: 139) - Starting new node TemplateProtocol
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 0
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 1
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 2
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 3
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 4
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 5
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 6
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 7
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 8
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 9
1 : (           platform.(*Localhost).Cleanup:  92) - Nothing to clean up
1 : (             platform.(*Localhost).Start: 161) - Starting 8 applications of simulation TemplateProtocol
1 : (                       platform.Simulate: 117) - Started counting children with timeout of 1000
1 : (                       platform.Simulate: 121) - Found all 21 children
1 : (                       platform.Simulate: 139) - Starting new node TemplateProtocol
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 0
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 1
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 2
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 3
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 4
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 5
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 6
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 7
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 8
1 : (          main.(*SimulationProtocol).Run:  83) - Starting round 9

This shows that it started 8 conodes (they, however, just running inside of goroutines; you will not see them as separate processes in a ps -ef listing). This is set from "Servers = 8" in the protocol.toml file.

There are two runs, once for depth = 1 and once for depth = 3. For depth = 1, and branching factor 4, there are 5 nodes expected (1 root + 4 children = 5). Once all of the 10 rounds are finished, the conodes are shut down. The simulation continues for depth = 3, with 1 + 4 + 4*4 = 21 children expected. Because 21 children > 8 conodes, some of the conodes will be hosting 3 children.

You could add the -debug 3 argument and get much, much more debug output from each conode showing what it is seeing and doing.

Running simulations with "go test"

The simul_test.go file shows that simulations can be launched from within standard Go tests. This would make it possible to have different tests that use different toml files in order to test different sizes of networks, etc.