N-body simulations are now the standard approach to studying non- linear structure formation. With ~trillion particle simulations now possible, containing many billions of dark matter halos, measures of key quantities (such as the halo mass function) can be made at extremely high precision. The hierarchical nature of CDM means that most of those halos are represented by only a relatively small number of particles however. I will describe how this "small-N" affects the results of N-body experiments in systematic ways, and describe other problems which arise when using N-body simulations to simulate the next generation of galaxy surveys. I will also describe work underway to circumvent these problems to allow measurements to be made from N-body simulations not only with high precision, but also with high accuracy.