EN
Back to the archive

The encyclopedia · Engineering & Operations · Technical decision · 1992–2000

NetApp's WAFL made backups a pointer, so snapshots became free and instant

WAFL never overwrites blocks: a snapshot is a saved root pointer, giving instant, space-efficient backups that made NetApp the NAS leader.

NetApp (Network Appliance)

The solution

In 1992 Dave Hitz, James Lau and Michael Malcolm founded Network Appliance to build a dedicated NFS file server. Nightly backups of Unix file systems meant hours of copying or tape, and after a crash, consistency checks could take as long as the downtime itself.

Their 1994 USENIX paper introduced WAFL: the file system writes anywhere and keeps the old root in place, so a Snapshot is a read-only clone created in an instant by copying a pointer. Copy-on-write means snapshots consume space only as blocks diverge; WAFL keeps up to 20 scheduled snapshots online and exposes them through a .snapshot directory so users can recover deleted files themselves.

The same mechanism eliminated long consistency checks after unclean shutdowns, a key selling point for the appliance. NetApp's Snapshot feature became an industry standard; the company IPO'd in 1995, grew past $1 billion in revenue and became the leading NAS supplier.

Why it worked

  • A snapshot is a pointer copy, so creating it takes milliseconds, not hours.
  • Copy-on-write means old versions cost almost no disk until data changes.
  • Users recover their own deleted files via the .snapshot directory.
  • Fast restart after crashes removed the worst pain of file servers.
What it achievedNever overwrite; snapshot is a pointerinspired

What can be applied

If the thing you dread (backup, undo, restart) is expensive, change the write path: when every change is versioned by construction, the recovery artifact is a pointer you already have.

Aftermath

Snapshots and their descendants (SnapMirror, FlexClone, deduplication) remained NetApp's product core for three decades. The 1994 paper is a classic of storage systems, and every major file system since — including ZFS and Btrfs — borrows its copy-on-write snapshot idea.

Sources

spotted an error? The archive wants to know.

Related cases