The Filesystem Landscape
concept"Which filesystem is best?" is the naive question, and the internet will happily fight about it forever. Actually, filesystems are tools for different jobs, and the real question is "what does each one promise?" Most of the familiar ones promise: your records will survive a crash. ZFS promises something stronger and rarer: your data will be verified, and if a copy exists, repaired—across many disks, as one system. Nothing else in wide use makes that full promise; one system (Btrfs) makes most of it. That difference, not speed or popularity, is what you're choosing when you choose ZFS.
The roster, one honest line each:
- ext4 — the Linux default. Journaled, fast, two decades of debugging. Trusts the disk completely. The right answer for most single-disk Linux machines.
- XFS — ext4's heavy-duty sibling (the default on Red Hat systems); excels at large files and parallel work. Same trust in the disk.
- NTFS — Windows' filesystem since the 90s. Journaled, capable, and not really a choice—it's what Windows is.
- APFS — Apple's modern filesystem. Shares real DNA with ZFS (copy-on-write, snapshots) but skips data checksums and multi-disk redundancy: it's the single-SSD-in-a-laptop distillation of the ideas.
- Btrfs — the closest relative: ZFS's convictions (checksums, snapshots, multi-disk pools) rebuilt inside the mainline Linux kernel. Mature for single disks and mirrors; its RAID5/6-style parity modes never became trustworthy, which is the caveat that keeps it second.
- exFAT / FAT32 — the interchange formats. No journal, no ownership, no promises; readable by everything. Perfect for a camera card, unthinkable for your archive.
What ZFS uniquely combines—each piece exists somewhere else; the combination is the offer:
- Verification (checksums on all data) — Btrfs yes; ext4/XFS/NTFS/APFS no.
- Repair from redundancy (self-healing across disks) — Btrfs partially; nothing else even attempts it.
- Whole-stack scope (one system managing disks, redundancy, and files) — traditionally three separate tools everywhere else.
- Snapshots and replication as first-class verbs — APFS and Btrfs have snapshots; ZFS's send/receive replication is the piece the others lack or approximate.
And the honest costs, so the comparison isn't a sales pitch: ZFS wants memory (its cache is a feature, not a leak), lives outside the mainline Linux kernel for licensing reasons (an extra installation step, not a quality verdict), and asks you to learn a real mental model—which is, after all, why this site exists. When the machine is one laptop with one SSD, ext4 or APFS is genuinely the better fit. When the machine is where your data lives, the combination above is the argument.