d-graid · research · also Degrades Gracefully RAID

D-GRAID

D-GRAID is a research storage-array design that combines fault-isolated placement and selective metadata replication to preserve access to surviving files after excess failures.

The FAST 2004 Alexander prototype supported ext2 and VFAT with RAID-0 or RAID-1 user-data layouts; RAID-5 support was future work. Replicated naming metadata keeps surviving files reachable, while additional striped copies recover parallelism lost by fault-isolated placement. [1]

Its live-block recovery conservatively includes possibly live blocks because delayed filesystem updates can make on-disk allocation information stale. This is a filesystem-aware prototype, not OpenZFS dRAID or a generic degraded-array state. [1] [2]

Relations

related
Degraded array, RAID rebuild, dRAID (OpenZFS), Distributed spare

Limitations

  • Partial availability depends on surviving metadata and file placement; it does not guarantee that every file survives. [1]

Citations

  1. Sivathanu, Prabhakaran, A. C. Arpaci-Dusseau and R. H. Arpaci-Dusseau. Improving Storage System Availability with D-GRAID. FAST 2004, sections 3, 5-7 and 9. (accessed 2026-11-09)
  2. OpenZFS 2.3 zpoolconcepts(7): dRAID virtual devices and hot spares. (accessed 2026-11-09)

Further reading

  1. Hesela: partial file availability, dRAID and recovery milestones (accessed 2026-11-09)

JSON: /data/terms/d-graid.json

Published by Hesela team · Updated

Report a correction · Editorial standards · Reuse guidance

Cite this page
Hesela team. D-GRAID. Hesela. 2026-11-09. https://hesela.dev/terms/d-graid/