Information Technology Reference
In-Depth Information
System
architecture
System
tests
Simulation engine
Single classes
and methods
Software
design
Unit
tests
Decomposition
and design
Realization, test
and integration
Implementation
Figure 9.11:
Development of the example implementation follows a
simplified V-Model.
citly, but indicated via the function name
informed()
(instead of
uninformed()
).
9.3.4 Parallelization on cluster-level
Considering parallelization on cluster-level the implementation of a
simulation engine is considerably more complex. This is largely due
to the fact that any simulation engine which employs cluster-level
parallelization is a decentralized software system with all its benefits
and pitfalls. The most important choice to make when implementing
such a simulation engine is to decide whether a central control node
exists or not. On the one hand, a central control node simplifies
many aspects (e. g., management issues like starting and stopping a
simulation and collecting simulation results). On the other hand, a
central control node may easily become a bottleneck and thus limit
optimal performance speedups.
9.4 Test and verification
As one of the main intentions of the example implementation is to
provide a reference (i. e., to serve as a guideline for implementing
simulation engines), the example implementation follows the GRAMS
reference model closely. Development of the example implementation
