As the number of components in large supercomputers grows, so does the possibility of component failure