ToolScore

OpenZFS

OpenZFS · cloud storage · reviewed by Stan K. · Source: data snapshot · data as of 2026-11-12

The layer underneath it all — checksums, snapshots, RAID-Z. It never speaks S3; it is what your NAS or object gateway stands on.

#6 of 12 in Cloud Storage

Try OpenZFS for $0
Real price
$0 · OSS
open-source
Free tier
Yes
Free and open source under CDDL. The bill is electricity, hardware and on-call time — self-hosting trades the
Limits
7.0/10
$0 · CDDL · filesystem layer (zfs-2.4.4, 21.08.2026)
Verdict
NICHE
No plan limits — it is a filesystem, not a service. Delivers checksums, snapshots, RAID-Z. It never speaks S3;

ToolScore breakdown — why 64/100

Price / Value
6.5
Token & Rate Limits
7.0
Speed / Latency
6.0
Privacy & Retention
7.0
BYOK (Own API Key)
3.0
Output Quality
7.5
Stability
8.0
Ease of Setup
6.0
Billing Transparency
7.0
Support & Community
5.0

10 criteria × 10 points. Low pillars are visible on purpose — that's the point of a score you can audit. Methodology →

What it does well

  • ▸Checksums, snapshots, RAID-Z — the integrity foundation everything above stands on
  • ▸zfs-2.4.4 shipped 21.08.2026 — steady maintenance on the base layer
  • ▸CDDL: a per-file weak copyleft with no network-service obligations
  • ▸Composable: it is what your TrueNAS or your object gateway stands on, whatever S3 server you pick

Where it costs you

  • ▸It never speaks S3 — no API, no protocol, no service semantics
  • ▸ARC cache defaults to ~50% of system RAM — set an explicit limit on small hosts
  • ▸Avoid dedup — the hub's one-word warning for a feature that eats RAM for breakfast
  • ▸Scored strictly as an object-storage pick it lands low (64): it is the layer beneath, not the store

Hidden traps — the stuff the pricing page doesn't say

medium riskARC cache defaults to ~50% of system RAM

Out of the box, OpenZFS's ARC cache will claim roughly half of system RAM. On a small host sharing duties with containers or an S3 gateway, that default starves everything else — set an explicit limit on small hosts. And avoid dedup: the RAM cost is the trap.

low riskA filesystem is not a storage service

OpenZFS appears in the hub's table because everything else stands on it — TrueNAS runs on it, object gateways store on it. But it never speaks S3: no endpoint, no protocol. Buying a disk and creating a pool is the start of a storage stack, not the end.

Try OpenZFS

Know the limits before you pay: read the traps above, then decide with eyes open.

Visit official site

Affiliate link where available — never affects the score.

What the advertised price hides
Free and open source under CDDL. The bill is electricity, hardware and on-call time — self-hosting trades the provider egress bill for your own network and on-call time. Current: zfs-2.4.4 on 21.08.2026. The hub lists it as the filesystem layer — the foundation, not a storage service.
Sources & freshness

Data as of 2026-11-12. Prices and limits change monthly — report anything stale. How we verify: methodology.

Closest alternatives

OpenZFS review

In-depth analysis · last updated 2026-11-12

The floor of the stack

OpenZFS is in the hub's object-storage table for taxonomy, not competition: it is the layer underneath it all — checksums, snapshots, RAID-Z — and it never speaks S3. Your TrueNAS runs on it; your object gateway stores on it. The hub scores it strictly as an object-storage pick, which is why it sits mid-table rather than at the top: it is the foundation, not the store.

What it gives every row above

The integrity primitives that make self-hosted storage defensible are here: checksums that catch silent corruption, snapshots that make deletes survivable, RAID-Z that makes disk loss routine. zfs-2.4.4 shipped 21.08.2026 — steady maintenance on the base layer. CDDL is a per-file weak copyleft: no AGPL-style obligations at the service boundary, which is why everyone from appliances to gateways builds on it freely.

The two tuning sentences

The hub's field advice fits in two lines: ARC cache defaults to ~50% of system RAM — set an explicit limit on small hosts. And avoid dedup. Both traps are RAM-shaped; both bite hardest exactly where self-hosters run — small boxes doing several jobs at once.

How the pillars land

Track record 8: zfs-2.4.4 on 21.08.2026 — the base layer of a large ecosystem, maintained on schedule. Feature set 7.5: checksums, snapshots, RAID-Z — the deepest integrity toolset in the table. Real limits 7 and Privacy 7: your disks, your filesystem, no meter and no third party. Pricing transparency 7: $0 under CDDL with the traps documented in the open. Price/Value 6.5 and Ease 6: free but expert-tuned — ARC limits and dedup avoidance are not optional knowledge. Transfer speed 6, neutral: no object-transfer facts apply. S3/API interop 3: it never speaks S3 — the category's core requirement, absent by design. Support & docs sits at the neutral 5 — nothing SLA-like verifiable in this snapshot.

Verdict — NICHE, 64/100

You do not choose OpenZFS instead of an object store — you choose it underneath one. Set the ARC limit, skip dedup, and let SeaweedFS or Garage speak S3 on top of the pool.

From the insights