Information Technology Reference
In-Depth Information
3 Parallel and distributed
multi-agent simulation
This chapter gives an overview of basic terminology in the area of
parallel and distributed multi-agent simulation and presents existing
approaches and their limitations for parallel execution of multi-agent
simulations.
3.1 Basic terminology
According to [36], the four main reasons to consider parallel and
distributed execution of a simulation are:
1. Reduced execution time.
2. Geographical distribution.
3. Integrating simulators and simulation models that execute on
machines from different manufacturers.
4. Fault tolerance.
As shown in Figure 3.1, hardware platforms providing multiple pro-
cessors are divided broadly into two categories. Parallel computers
are usually homogeneous machines which are connected by special-
ized interconnects to reduce latencies between the machines. On the
other hand, distributed systems consist of geographically distributed,
heterogeneous machines which are usually loosely coupled (sometimes
only temporarily for a specific purpose).
The distinction between parallel and distributed computers is not
clear-cut. A cluster or network of workstations may be classified
