{
  "slug": "parabit",
  "term": "ParaBit",
  "aliases": [],
  "definition": "ParaBit is a research design that performs bitwise operations inside NAND flash by controlling sensing sequences and page-buffer latches.",
  "note": "Placement: Basic ParaBit pairs operand bits in the same MLC cell. Its location-free variant uses aligned cells on one bitline; the described AND and XOR sequences require the second operand in an LSB page. XOR support adds circuitry. Location-free does not mean arbitrary page placement. [parabit-paper]\n\nError handling: ECC runs during operand reallocation, but cannot correct new sensing errors during the subsequent in-flash logic. The design disables scrambling and encryption for prepared operands. Flash-Cosmos later targets both serial sensing and raw-error exposure through multi-wordline sensing and enhanced SLC programming. [parabit-paper] [cosmos-paper]\n\nEvidence: The MICRO 2021 paper combines SSD simulation with measured result-transfer times and a separate real-MLC error study. This is not an end-to-end commercial ParaBit SSD benchmark. Its custom use of reserved NVMe command fields is a research interface, not standard NVMe capability discovery. [parabit-paper]\n\nOperation order: In a specified Hamming(7,4) model, fault-free AND followed by decoding gives 96 wrong payloads out of 256 ordered pairs; all raw payloads were correct. Extended Hamming(8,4) rejects 168 pairs instead. These exact toy-code counts are not NAND failure rates. [ecc-order]",
  "status": "research",
  "advantages": [
    {
      "text": "Combining operands before transfer can reduce traffic between NAND, the controller and host.",
      "citationIds": [
        "parabit-paper"
      ]
    }
  ],
  "limitations": [
    {
      "text": "Operand relocation adds writes and can dominate execution time; placement and error protection must be costed together.",
      "citationIds": [
        "parabit-paper"
      ]
    },
    {
      "text": "Conventional ECC and randomization cannot simply be applied unchanged to arbitrary AND/OR computations on stored representations.",
      "citationIds": [
        "cosmos-paper"
      ]
    }
  ],
  "related": [
    "flash-cosmos",
    "computational-storage",
    "nand-flash"
  ],
  "supersedes": [],
  "supersededBy": [],
  "firstSeen": null,
  "lastSeen": null,
  "citations": [
    {
      "id": "parabit-paper",
      "title": "Gao et al. ParaBit. MICRO 2021, pp. 59-70, sections 4.2-4.4, 5.1 and 5.8. DOI: 10.1145/3466752.3480078.",
      "url": "https://par.nsf.gov/servlets/purl/10316348",
      "accessed": "2026-11-18"
    },
    {
      "id": "cosmos-paper",
      "title": "Park et al. Flash-Cosmos. MICRO 2022, sections 3.2, 4, 5 and footnote 11. DOI: 10.1109/MICRO56248.2022.00069.",
      "url": "https://arxiv.org/pdf/2209.05566v1",
      "accessed": "2026-11-18"
    },
    {
      "id": "hesela-nvme",
      "title": "Hesela: NVM Express, in-flash computation and error-protection boundaries.",
      "url": "https://hesela.com/tech/nvme/",
      "accessed": "2026-11-18"
    },
    {
      "id": "ecc-order",
      "title": "Hesela: ECC operation order, exact Hamming-code construction, method and results.",
      "url": "https://hesela.dev/research/ecc-operation-order/README.md",
      "accessed": "2026-11-18"
    },
    {
      "id": "ecc-order-replay",
      "title": "Hesela: offline ECC operation-order replay, Python code, tests and exact results (ZIP).",
      "url": "https://hesela.dev/research/ecc-operation-order.zip",
      "accessed": "2026-11-18"
    }
  ],
  "updated": "2026-11-18",
  "provenance": [
    "Literature synthesis from the original MICRO 2021 paper and the later Flash-Cosmos study; no hardware replication or new performance measurement.",
    "The original error experiment reports nonzero errors. No universal bit-exact guarantee, present-day product support or cross-generation error rate is inferred.",
    "The ECC operation-order example enumerates synthetic finite populations with ideal logic and fixed layouts, not physical NAND failures. [ecc-order]"
  ]
}
