Technical Reference
Core concepts, glossary, and ecosystem readiness.
What is FDP?
Flexible Data Placement is an NVMe standard feature that assigns Placement ID (PID) tags to write requests, directing the SSD to physically separate data with different lifetimes and access patterns. This prevents data mixing within erase blocks, reducing Garbage Collection overhead and Write Amplification Factor.
Why FDP Matters
SSDs use flash memory, which cannot overwrite data in place. GC must relocate valid data and perform additional writes to reclaim space, causing WAF to rise. When short-lived and long-lived data share the same erase blocks, GC overhead compounds—degrading both performance (especially p99 latency) and drive endurance. FDP allows the host to proactively tag writes with group identifiers, guiding the SSD to physically separate data at the erase block level itself.
FDP vs Hot/Cold Tiering
| Hot/Cold Tiering | FDP | |
|---|---|---|
| Approach | Reactive data migration | Proactive write-time tagging |
| Granularity | Cache layer level | Erase block level |
| GC Control | Indirect | Direct physical separation |
| Classification | Binary (hot/cold) | Multi-class (DataFount: 6-class) |
“FDP-Based SSD Slicing: A logical partitioning technique in which the host assigns PID tags to I/O operations using the NVMe FDP feature, directing the SSD to route data into distinct Reclaim Unit pools.”
Input Feature Reference
A: I/O Pattern Features
B: Data Semantic Features
C: System/Policy Features
Group C features are for mature policy engine integration only. MVP requires Groups A and B.