Concluding remarks
The development and implementation of anMPI-enabled
GA is discussed. Due to the very nature of genetic algorithms,
linear speedup is possible with minimal effort –
a good example of an embarrassingly parallel problem.
The developed algorithms work well in both homogenous
and heterogeneous network of workstations clusters. Such
an environment can include shared as well as distributed
memory. Due to the nature of the problems discussed
in this paper and genetic algorithms, the heterogeneous
cluster using COTS components performed as well as the
more expensive homogenous cluster.