Employing domain replication with the Neumann-Ulam method will be an
exercise in batch statistics. By using replication, we also expect to incur an
additional memory overhead. As noted in previous work, domain replication
was coupled with domain overlap in order to improve load balancing and
scalability in the implementations. As replication will likely be required to
solve future resiliency issues and may even by required by this work in order to achieve scalability, its parallel performance must also be measured. We
will therefore measure the memory and performance overhead incurred by
varying the amount of domain replication leveraged by the Neumann-Ulam
solver with respect to a fixed linear system.