Distributed external simulations

At the moment this section applies only to MLDesigner on Linux.

The ability to run a simulation extern on a number of computers simultaneously is useful where complex simulations with numerous iterations result in long simulation run times.

The computers used for the distributed simulation must all have the same architecture and the same version of MLDesigner otherwise the simulation will not run.

The first step towards running a distributed simulation is to ensure that all workstations have MLDesigner installed in a shared environment. The $MLD_USER environment variable should also be set to a shared environment, that is, the command mld should start the same MLDesigner on all computers. The computers to be used must also be registered in the /etc/hosts.equiv file.

Host registry for distributed simulations

Workstation

/etc/hosts.equiv entries

Computer 1

Computer2

 

Computer3

Computer 2

Computer1

Computer 3

Computer1

On computer 2 and 3 the rshdaemon must be running. To test if this is the case type:

rsh Computer2 rsh Computer1 hostname

This command should return Computer1.

In the Used Machines property of the Simulation Properties window enter the computers you wish to use. If this field is empty the simulation runs only on localhost. A list of servers can be added to the Used Machines property by editing the ~/.mld/.mldrc file. By default localhost is entered in the <SHARED_SERVERS> environment. To add servers to this list, type the server name next to or below the localhost entry. Save the file and restart MLDesigner. The servers entered now appear in the Used Computers list when MLDesigner is in Simulation Mode.

The simulation can now be run. Note that the rshscheduler runs each iteration parallel although only one iteration per computer can be run at a given time. If the progress bar is visible it is possible to observe the simulation as it is executed on all computers.