As memory requirements grow, and advances in memory technology slow, the availability of sufficient main memory is increasingly the bottleneck in large compute clusters. Can far memory improve job throughput? Checkout the paper and presentation.