A visualization and steering application, GAVis, has been developed to aid in understanding the behavior of and guiding the convergence of genetic algorithms running in parallel over long time periods. When classification techniques such as Support Vector Machines (SVMs) paired with complete leave-one- out validation are used as a fitness function for identification of markers in - omic data, the time to complete one generation can exceed an hour on modern high-performance computing clusters. Separate solution populations on "islands" can help maintain a more diverse solution space and conveniently map to compute nodes on a cluster. Adjustments can be made at runtime to speed the convergence of genetic algorithms by stimulating lagging island populations with migrations of high-performing individuals or by selectively increasing mutation rates.