In this section we study the performance of le I/O as A
aggregator nodes write to the parallel le system. As in the
network study, we scale both load and core counts in order to
evaluate different aspects of the storage system. We present
excerpts from our study that exhibit interesting behavior.