{
  "status": "research",
  "advantages": [],
  "limitations": [],
  "supersedes": [],
  "supersededBy": [],
  "firstSeen": null,
  "lastSeen": null,
  "updated": "2026-11-16",
  "slug": "vstream",
  "term": "vStream",
  "aliases": [
    "virtual stream management"
  ],
  "definition": "vStream is a research SSD design that groups host-assigned virtual streams into fewer physical streams using observed data lifetimes.",
  "note": "Mapping: The controller estimates lifetime from pages invalidated by overwrite or DISCARD, then periodically reclusters streams. Applications still supply virtual IDs; the scheme does not infer file semantics from ordinary writes. [paper]\n\nEvaluation: Modified Linux 4.10.0 and RocksDB 5.8.0 produced traces for the DROLL SATA SSD simulator: 128 GB, 7% over-provisioning, ten virtual streams and four physical streams. Compression was disabled. [paper]\n\nResults: The RocksDB experiment reports 70% higher throughput and 18% fewer valid pages copied during GC than its no-stream baseline. The latter counts relocated pages; it is not an 18% reduction in device byte write amplification. Neither result is a physical-drive measurement. [paper]\n\nInterpretation: An average of already-invalidated pages excludes still-live data. It should not be treated as the complete lifetime distribution of a changing workload.",
  "related": [
    "multi-stream-ssd",
    "fstream",
    "ftl"
  ],
  "citations": [
    {
      "id": "paper",
      "title": "Yong, Jeong, Lee and Kim. vStream: Virtual Stream Management for Multi-streamed SSDs. HotStorage 2018, sections 4-6.",
      "url": "https://www.usenix.org/system/files/conference/hotstorage18/hotstorage18-paper-yong.pdf",
      "accessed": "2026-11-16"
    },
    {
      "id": "record",
      "title": "USENIX HotStorage 2018: vStream publication record",
      "url": "https://www.usenix.org/conference/hotstorage18/presentation/yong",
      "accessed": "2026-11-16"
    }
  ],
  "provenance": [
    "Full-paper source synthesis, not a benchmark reproduction. Physical-SSD implementation is future work in the 2018 paper. [paper]",
    "The lifetime-distribution qualification is an interpretation of the estimator's observation boundary, not a measured bias result."
  ]
}
