TrueNAS Scale vs Unraid: Choosing a NAS OS
Two mature, opinionated NAS operating systems with opposite ideas about disks — and how to tell which one fits the pile of hardware you actually have
Contents
When someone asks me which NAS operating system to run, the honest answer starts with a question back: what does your pile of disks look like, and how do you feel about buying more of them in matched sets? Because the two mainstream choices in late 2023 — TrueNAS Scale and Unraid — are both excellent, both mature, and built on genuinely opposite philosophies about how disks should be arranged. Pick the one whose philosophy matches your hardware and habits, and the thing is a joy. Pick against your grain and you will fight it forever. This is the comparison I wish someone had laid out for me before I built my first proper NAS, framed around the decisions that actually matter rather than a feature checklist.
The one difference that drives everything: how they treat disks
Start here, because every other trade-off flows from it.
TrueNAS Scale is ZFS, and ZFS thinks in pools of vdevs. You group disks into a virtual device — a mirror, or a RAIDZ1/RAIDZ2 group — and those disks are bound together with real-time redundancy, block-level checksums, and self-healing. Data is striped across the vdev, so all its disks work together and, in practice, want to be the same size, because a vdev is sized to its smallest member. In exchange you get the best data-integrity story in the business: every block is checksummed on read, and with redundancy ZFS repairs silent corruption automatically before it reaches you. Snapshots are instant and free, and replication to another box with zfs send is built into the platform’s DNA — the send/receive time-travel workflow is a first-class feature of the platform rather than something you graft on afterwards.
Unraid is a flexible parity array, and it thinks in independent disks. Each data disk is its own standalone filesystem, and one or two dedicated parity disks protect the lot. Disks can be any size you like, with the single rule that a parity disk must be as large as the biggest data disk. Files live whole on individual disks, so only the disk being read spins up, and a disk pulled from the array remains readable on its own in any machine. Expansion is trivial: slot in another disk of any capacity and the array absorbs it. If you have read my piece on SnapRAID and MergerFS, Unraid’s array is essentially that same flexible-disk philosophy productised into a polished, integrated OS with a licence fee attached.
That contrast decides most of the outcome. ZFS gives you integrity and performance at the cost of rigidity and matched disks. Unraid gives you flexibility and cheap expansion at the cost of ZFS’s self-healing striped resilience for the array itself.
Expandability: the decision that will haunt you
This is where people most often regret their choice, so weigh it carefully against your future self.
Unraid wins outright on incremental growth. Bought one 12 TB disk this month and a 16 TB next year? Slot each in as it arrives. The array does not care that your disks are a museum of different capacities, which is exactly the drawer-of-mismatched-disks reality most homelabs live in.
ZFS in December 2023 is far more demanding here. You expand a pool by adding whole vdevs, so growing a RAIDZ pool has historically meant adding another complete RAIDZ group of matched disks, rather than slotting in a single drive. The ability to expand a single RAIDZ vdev by one disk is on the horizon in OpenZFS but not yet shipping in a release, so plan your ZFS pool topology as though you cannot cheaply widen a vdev, because right now you cannot. You can replace every disk in a vdev with larger ones one at a time, resilvering between each, to grow capacity — slow, but it works. The practical upshot: with ZFS you design the layout up front and buy disks in sets; with Unraid you grow one disk at a time forever.
Data integrity and failure behaviour
ZFS is the stronger story here, and if your data’s integrity is paramount it may settle the question on its own. Checksums on every block mean ZFS detects silent corruption, and redundancy means it repairs it during a scrub, self-healing before bad data reaches your application. Unraid’s parity protects against a disk dying, and its 6.12 release added the option of ZFS-formatted pools for cache and app data, but the classic parity array does not checksum and self-heal the way a ZFS vdev does.
Failure behaviour differs in a way worth internalising. Lose more disks than your redundancy allows in a ZFS vdev and the whole pool is gone, because the data was striped across it — the same brittle-at-the-edge property I dug into in rethinking RAID when mirrors beat parity in a homelab. Exceed Unraid’s parity and you lose only the data on the disks that actually died, because every surviving data disk is still a whole, independent, readable filesystem. For bulk media where losing a couple of disks’ worth of films is survivable, Unraid’s degradation is gentler. For a dataset where you need the whole thing or nothing, ZFS’s integrity and redundancy model is what you want.
Apps, containers and VMs
Both run Docker containers and KVM virtual machines, so both make fine homelab application hosts, but the flavour differs.
Unraid’s Community Applications plugin is the friendliest container experience in this class. It is a searchable catalogue of templates that fill in sensible defaults, and for someone who wants to click “install” on a media server, a download client and a dashboard, it is hard to beat. GPU passthrough for transcoding or a gaming VM is well-trodden and heavily documented in the Unraid community.
TrueNAS Scale runs apps through a Kubernetes layer (k3s) with its own catalogue, plus the ability to run standard Docker images. It is more powerful and more structured, and it is fussier — the Kubernetes underpinning means more moving parts and occasionally more opaque failures when an app misbehaves. Both run VMs via KVM competently. For pure application ergonomics in 2023, Unraid feels lighter; for an integrated ZFS-plus-apps appliance from one vendor, TrueNAS is coherent.
Hardware, RAM and the ECC question
ZFS likes RAM. It uses free memory as an aggressive read cache (ARC), and while the old “1 GB per TB” rule was always more folklore than requirement, a ZFS box is happier with 16 GB or more, and deduplication (which you should mostly avoid at home) is genuinely memory-hungry. ECC RAM is strongly recommended for ZFS — the filesystem protects your data on disk brilliantly, and ECC protects it in memory before it is written, closing a gap that non-ECC leaves open. ZFS runs fine without ECC, and plenty of homelabs do, but the pairing is the belt-and-braces the design assumes.
Unraid is lighter on RAM for the array itself and runs comfortably on modest hardware, which makes it attractive for repurposing an old desktop or a low-power build. It boots from and runs on a USB flash drive — the licence is tied to that drive’s GUID — which keeps your disk bays free for storage. Both are happy on typical homelab hardware; ZFS simply rewards more RAM and ECC where Unraid does not ask for them.
Licensing and cost
This is a genuine difference worth weighing, however dull licensing sounds. TrueNAS Scale is free and open source, developed by iXsystems, with no licence fee at any scale. Unraid is commercial, a one-time purchase with tiers historically priced by the number of attached storage devices, plus a free trial to evaluate it. Neither cost is large in the context of a NAS full of disks, but “free and open source forever” versus “a one-off licence tied to your boot USB” is a values question for some people, and worth naming honestly rather than glossing over.
A concrete example: laying out storage
To make the philosophies concrete, here is how you might create a mirrored ZFS pool from the TrueNAS shell, the sort of layout I would use for data I care about — two mirror vdevs striped together for capacity and integrity:
| |
Notice what this bakes in: four disks committed up front, in matched pairs, arranged so you understand the redundancy before a byte is written. The Unraid equivalent has no comparable command because its model is the inverse — you assign each disk a role (data or parity) in the web UI and add more whenever you like, with none of the up-front topology planning ZFS demands. That difference in ceremony is the difference in philosophy made visible.
Troubleshooting and the traps to avoid
ZFS pool built on mismatched disks wasting capacity. A vdev sizes to its smallest disk. Put a 4 TB and an 8 TB in a mirror and you get 4 TB. Match disks within a vdev, or choose Unraid, whose whole point is not caring.
Assuming you can grow a RAIDZ vdev by one disk today. You cannot, in a shipping release as of late 2023. Design the pool for the capacity you will need, or use mirror vdevs (which you can add in pairs) if you want a smoother growth path.
Unraid parity disk smaller than a data disk. The array refuses, because parity must cover the largest member. Your biggest disk sets the floor for the parity disk — buy accordingly, and buy the parity disk first when upgrading capacities.
Treating either array as a backup. Both give you redundancy against disk failure on one machine, and redundancy is not a backup. A fire, a theft, a bad rm, or ransomware takes the whole NAS regardless of parity or RAIDZ. You still need an independent, off-machine copy — ZFS replication makes that easy on TrueNAS, and either platform can push backups off-site.
Non-ECC ZFS on a memory-flaky box. ZFS trusts what it is handed in RAM. Bad memory can corrupt data on its way to a checksummed disk, and the checksum will faithfully record the corruption. Test memory thoroughly, and prefer ECC for a ZFS build if you can.
Verdict: which one, for whom?
Reach for TrueNAS Scale when data integrity is your top priority and you are willing to plan disk layout in matched sets: important documents, photos you cannot re-shoot, anything where silent corruption would be a disaster. Its ZFS foundation, free self-healing, instant snapshots and built-in replication are the strongest integrity story available, it costs nothing, and it rewards a machine with plenty of ECC RAM. The price you pay is rigidity in expansion and a slightly fussier apps layer.
Reach for Unraid when your reality is a drawer of mismatched disks you want to grow one at a time, your bulk data is media where losing a disk’s worth is survivable, and you value the friendliest container experience and low-ceremony expansion over ZFS’s integrity guarantees. It sips RAM, runs on modest hardware, and its app catalogue is the gentlest on-ramp in the category. The price you pay is a licence fee and a weaker self-healing story for the main array.
Plenty of people run both over time, or run Unraid with ZFS-formatted pools to split the difference. There is no wrong answer here, only a mismatch between an OS’s philosophy and your disks. Work out which describes your hardware and your habits, and the choice mostly makes itself — then, whichever you pick, go and build the off-machine backup that neither one replaces.
Related Content
