Iron Mountain: It's Time to Talk About Hard Drives

Original Article ↗ Hacker News Discussion ↗

Hard drives, once seen as a safe long‑term archive medium, are proving alarmingly fragile: a professional storage firm reports that roughly 20% of 1990s-era disks it receives are now unreadable. Commenters contrast this with tape, optical media, and cloud storage, arguing that no medium is truly permanent and that reliable preservation depends on active processes: redundancy, periodic integrity checks, and regular migration to new formats and technologies. The broader concern is that institutions and individuals alike often underestimate costs, overlook dependencies (file formats, software, interfaces), and may ultimately lose culturally or legally important data without ongoing stewardship.

Tape vs. Hard Drives for Archival

  • Many argue LTO tape “wins” for large-scale, long-term archival due to cheaper media, higher reliability, and clearer vendor roadmaps.
  • Others note tapes are effectively dead for home/very small-office use: drives are expensive, availability is poor, and consumer-friendly options vanished.
  • Critiques of HDD archival: powered‑down disks suffer mechanical failures (stiction, bearings), magnetic remanence loss, and interface/format obsolescence; 1990s disks are often unreadable.

Practical LTO Implementation

  • Several detailed posts describe DIY Linux/FreeBSD LTO setups:
    • Use LTO‑9 tabletop drives or autoloaders with SAS HBAs; watch connector and SAS‑generation compatibility.
    • Ensure end‑to‑end throughput to avoid “shoe‑shining” (underflowing the drive), e.g., via 10GbE and RAM disks/mbuffer.
    • Prefer simple tools (mt, dd, rsync) plus your own scripts over proprietary backup software; maintain a separate metadata/index database with hashes and possibly erasure codes (par2).
  • Autoloaders are recommended when staff time/remote data centers make manual tape swaps impractical; costs around mid‑four to low‑five figures are cited.

Media Longevity & Migration

  • Strong consensus: no medium is “store and forget.” All magnetic, optical, and solid-state media degrade or become unreadable due to:
    • Physical decay (bit rot, sticky‑shed, warped optical discs, SSD charge loss).
    • Obsolete drives, interfaces, and file formats/software dependencies.
  • Recommended approach: periodic verification (checksums/fixity checks) and remigration to newer media and formats every X years.
  • Some distrust of using HDDs powered off for years; others report success with disciplined rotation, ZFS/Btrfs scrubs, and regular checksum validation.

Cloud vs. Self-Hosted Archival

  • One camp: major cloud providers are the most practical long‑term option; they continuously replace failing disks and verify data, and multi‑cloud storage reduces risk.
  • Counterpoints:
    • SLAs are aggregate; they don’t guarantee safety of any specific dataset.
    • Legal/political/account-termination risks make sole reliance on cloud risky.
    • Costs (e.g., per‑TB per‑month cold storage) can exceed DIY HDD/tape for large archives.

Iron Mountain and Industry Practices

  • Iron Mountain is criticized for merely warehousing obsolete hard drives without active migration, leading to high unreadability rates.
  • Some defend it as historically a “vault for media” rather than a managed archive; recent pivots toward full data archiving/e‑discovery are noted.
  • Broader theme: archives are treated as cost centers, so many industries underinvest in proper processes despite cultural and legal importance.