Most caching and memory placement strategies still rely on simplified assumptions like uniform latency and mandatory caching. But modern systems are layered, unpredictable, and asymmetric. DRAM, SSD, and emerging CXL tiers behave differently, and legacy models can’t account for that complexity.