Features of Akaroa2©
Works on existing simulation programs
Any simulation program that can be linked with a library of C routines can be easily adapted to run under Akaroa2©. There is no need for any special parallel programming techniques.
Speedup approaching linear in the number of processors used
Processors running the simulation communicate only with a central controlling process and not with each other, so there is no synchronisation overhead to limit parallelism.
Real-time simulation data output analysis
Sophisticated and thoroughly tested statistical methods are used to continously analyse the precision of simulation output data as it is generated.
Automatic run length control
The user specifies the required precision and confidence level of the simulation results, and Akaroa2© automatically runs the simulation for long enough to produce those results. There is no need to guess the required run length in advance.
If processors are lost during a simulation, the simulation continues with the remaining processors. New processors can be added to a running simulation at any time, to replace lost processors or further speed up the simulation.
Portable to most variants of Unix
Implemented in the GNU dialect of C++, Akaroa2© has been tested at Canterbury under SunOS 4, Solaris 2 and Linux. Porting to other Unix variants is expected to be straightforward.
Can be used with other simulation packages
Akaroa2© can be used to control simulations built using other simulation packages. Interfaces currently exist for Ptolemy Classic, the Network Simulator NS2, and OMNeT++. Download the interfaces here.
Last modified: 23 Jun 2013