The rise of memory-intensive applications—from generative AI to hyperscale recommendation systems—has appeared in the era of big memory, demanding architectures that combine massive capacity with high performance. As traditional local memory struggles to meet these needs, systems now embrace far memory—including disaggregated memory pools and remote-server resources—to break capacity barriers and eliminate stranded memory. However, this expansion comes with tradeoffs. While big memory enables terabyte-scale working sets, the performance gap between local and far memory forces systems to employ transparent caching strategies, introducing read/write amplification overheads. The challenge lies in balancing this memory hierarchy to deliver both scale and speed for next-generation workloads. In order to efficiently address these challenges, it is necessity and important to redefine the design principle in the context of big memory. We believe that persistence is the first principle of big memory.