Granularity Bracket — FEMU between block & superblock GC
What it shows. The same conventional workload run three ways: EyanaSSDSim with block-granular GC, FEMU (which erases whole superblocks/lines), and EyanaSSDSim with superblock-granular GC.
The clincher. On prxy_0 the three values are
1.54 (block) ≤ 1.873 (FEMU) ≤ 2.14 (superblock) — FEMU is
bracketed by the simulator's two granularity modes. The ~40% swing from switching
erase granularity contains FEMU's value, proving that granularity — not a modeling
error — is what produces the WAF difference in the Validation graph.
Why granularity matters. Real NAND SSDs erase at superblock granularity for parallelism; a whole line is reclaimed together, so more valid pages are migrated → higher WAF. Block-granular reclaim can pick the single emptiest block → fewer migrations → lower WAF.
- prxy_0: block GC 1.54 ≤ FEMU 1.873 ≤ superblock GC 2.14. FEMU is bracketed.
- The block→superblock switch moves WAF by ~40% and contains FEMU's value, so granularity fully accounts for the sim–FEMU gap in the Validation graph.
- Superblock GC also raises wear concentration (DoEC 10.6 → 20.2 on prxy_0), the endurance cost of coarse-grained erase.
