How to Choose Between RAID 5, RAID 6, RAIDZ, and Mirrored Drives

RAID 5, RAID 6, RAIDZ, and mirrored drives are not just different names for the same safety feature. They make different tradeoffs between usable capacity, write behavior, rebuild risk, future expansion, and how stressful recovery becomes when a drive fails.

The right choice depends on the job. A small home NAS may be easier to live with as a mirror. A large HDD pool usually deserves dual parity. A VM or Docker-heavy server often benefits from mirrors. A media archive may care more about capacity. None of these layouts replaces backup.

The Wrong Question Is “Which RAID Is Best?”

The better question is: what are you trying to protect, and what are you willing to give up? RAID 5 gives more usable space but only one-drive fault tolerance. RAID 6 gives stronger protection but costs another drive of capacity. RAIDZ brings ZFS checksums and vdev design into the decision. Mirrors cost the most capacity, but they are simple, fast, and easier to expand in pairs.

Start with these questions before choosing any layout:

Question	Why It Matters
How many drives do you have?	2 drives usually means mirror. 5–8 large HDDs usually pushes toward dual parity.
Are the drives HDDs or SSDs?	Large HDD rebuilds are slow. SSD pools often care more about IOPS and latency.
Is this for media, backup, VM, Docker, or AI workloads?	Sequential storage and random app data prefer different layouts.
How painful is downtime?	Some layouts rebuild faster and keep performance more predictable.
Do you already have backup?	RAID without backup is still a single storage system that can fail.

If the answer is only “I want the most space,” RAID 5 or RAIDZ1 will look attractive. If the answer includes “I cannot afford panic during rebuild,” the choice usually changes.

Usable Capacity Is the Easy Part

Capacity is simple to calculate, but it is not the whole decision. A layout that looks efficient on day one may be harder to rebuild, expand, or trust once the pool holds years of photos, office documents, media libraries, Docker volumes, or AI project data.

Layout	Minimum Drives	Drive Failure Tolerance	Usable Capacity Pattern	Practical Meaning
Mirror / RAID 1	2	1 per mirror pair	About 50%	Simple, fast recovery, high capacity cost
Striped mirrors	4	Depends which drives fail	About 50%	Good performance and easier pair-based growth
RAID 5 / RAIDZ1	3	1	n-1 drives	Efficient capacity, narrow safety margin
RAID 6 / RAIDZ2	4	2	n-2 drives	Better for large HDD pools and important data
RAIDZ3	5+	3	n-3 drives	For very large or conservative storage pools

RAID 5 and RAIDZ1 look efficient because they only reserve one drive worth of parity. Mirrors look wasteful because half the raw capacity is used for redundancy. RAID 6 and RAIDZ2 sit in the middle: less efficient than single parity, but safer during drive failure and rebuild.

Single Parity Is a Budget Choice, Not the Safe Default

RAID 5 and RAIDZ1 are not automatically bad. They are just narrower fits than many beginners assume. They can make sense for small arrays, SSD-based pools, non-critical data, test systems, or media libraries that already have a separate backup.

The risk is simple: single parity can survive one drive failure. During rebuild, there is no extra drive-failure margin left. With large mechanical hard drives, a rebuild can take a long time, and the remaining drives are under sustained read pressure while the replacement drive is rebuilt.

Use RAID 5 or RAIDZ1 when capacity matters, risk is understood, and backup exists. Do not use it as the default answer for large HDD pools that hold irreplaceable data.

Dual Parity Is for Large Drives and Long Rebuild Windows

RAID 6 and RAIDZ2 give up another drive of capacity to survive two drive failures. That extra safety matters more as drive size grows, rebuild time increases, and the pool becomes more important.

For a 6-bay or larger HDD NAS, dual parity is often the practical baseline. It is especially relevant for family photo libraries, office documents, media archives, backup targets, private cloud storage, and multi-user NAS environments. The write path is heavier than single parity, but many NAS workloads care more about safety, capacity, and predictable recovery than raw write speed.

If you are building a larger ZimaOS storage system, the official ZimaOS ZFS setup guide is the right internal reference for creating a ZFS storage pool before layering shares, apps, or backup workflows on top.

Mirrors Cost More Capacity but Are Easier to Live With

Mirrored drives are expensive in usable capacity, but they are easy to understand. Each drive in a mirror pair contains a full copy of the data for that vdev. If one drive fails, recovery is more direct because the replacement is rebuilt from its mirror partner.

This is why mirrors are often preferred for VM, database, Docker app data, metadata-heavy workloads, and smaller personal servers. They are not always the most space-efficient layout, but they can feel better in daily use because random I/O and recovery behavior are more predictable.

Workload	Mirror Fit	Reason
VMs and databases	Strong	Random I/O and latency matter
Docker app data	Strong	Small writes, metadata, and databases benefit
Simple 2-drive NAS	Natural	Easy to understand and recover
Large media archive	Mixed	Capacity loss may be too high
Cold backup pool	Good but costly	Simple recovery, less capacity efficiency

A TrueNAS discussion on mirroring vs RAIDZ2 resilvering captures the real tradeoff: mirrors can resilver faster and with less pool-wide impact, while RAIDZ2 keeps an additional parity layer after one drive has failed. That is why the choice depends on workload, uptime needs, spare drives, and backup strategy.

RAIDZ Is Not Just RAID 5 With a Different Name

RAIDZ belongs to ZFS pool design. In ZFS, you are not only choosing a RAID level. You are choosing a vdev topology that shapes performance, redundancy, rebuild behavior, and future growth.

Klara’s explanation of OpenZFS vdev types and RAIDZ topology is useful because it frames ZFS pools from the top down. A zpool is built from vdevs, and those vdevs can be mirrors, RAIDZ1, RAIDZ2, RAIDZ3, or other device classes. If one top-level data vdev fails, the pool is at risk.

This is the part many users miss. With ZFS, the layout you create early can define the storage system for years. RAIDZ1, RAIDZ2, and mirrors are not just capacity choices. They are long-term topology choices.

Rebuild Risk Should Decide More Than Capacity

Most bad RAID decisions happen because capacity is easy to see and rebuild risk is invisible until a drive fails. A single-parity layout may look efficient, but the rebuild window is the moment when the pool is most exposed.

Mirrors usually rebuild more locally. RAIDZ layouts may involve more disks during recovery. RAIDZ2 gives you a second parity layer, which can matter if another drive misbehaves while the failed disk is being replaced. Neither approach is perfect, but each has a different failure shape.

Situation	Safer Direction
3 small SSDs with non-critical data	RAIDZ1 / RAID 5 can be acceptable
4 large HDDs with family data	RAIDZ2 / RAID 6 or two mirrors
6+ high-capacity HDDs	RAIDZ2 / RAID 6 minimum for most users
VM or Docker-heavy workload	Mirrors
Backup target with offsite copy	RAIDZ2 or mirrors, depending on capacity and restore speed
No separate backup	Fix backup first, then choose RAID

The layout should match how you want recovery to feel. If you would panic during a long rebuild, do not choose the layout only because it gives one more drive of usable space.

Expansion Plans Matter Before the Pool Holds Real Data

Expansion is where mirrors and RAIDZ feel very different. Mirrors are usually easier to grow in pairs. Add another mirror pair, and the pool gains capacity and performance from another vdev. That simplicity is one of the main reasons mirrors stay popular in home servers and homelabs.

RAIDZ expansion has improved, but it still deserves planning. The OpenZFS issue on OpenZFS RAIDZ expansion multiple-drive enhancement request highlights an important limitation: RAIDZ expansion can involve slow, serial expansion when adding drives one at a time, and users still need to think about performance impact and wear during the process.

Do not treat RAIDZ expansion as permission to choose the wrong parity level today. Expansion can add capacity. It does not turn RAIDZ1 into RAIDZ2, erase the need for backup, or remove the consequences of the original vdev design.

Match the Layout to the Real NAS Job

The best RAID layout is the one that matches the job of the NAS. Media storage, backup, Docker apps, VM storage, AI projects, and cold archive do not behave the same way.

NAS Job	Recommended Direction	Why
Simple 2-drive home NAS	Mirror	Easy, predictable, and simple to recover
4-drive family storage	RAIDZ2 / RAID 6 or two mirrors	Balance safety, cost, and rebuild behavior
6-bay media library	RAIDZ2 / RAID 6	Capacity efficiency with dual parity
VM / database / Docker host	Mirrors	Better random I/O and simpler recovery path
Backup-only NAS	RAIDZ2 or mirror	Choose based on restore speed and capacity need
AI NAS / RAG / app data	Mirrors for hot app data, RAIDZ2 for bulk storage	Separate active databases from large archive files
Cold archive	RAIDZ2 / RAIDZ3	Capacity and fault tolerance matter more than IOPS

For a compact always-on server, ZimaBoard 2 personal server fits simpler mirrored storage, backup, Docker, and light home server use. For larger multi-drive storage, media libraries, private cloud, and AI NAS workflows, ZimaCube 2 AI NAS is better aligned with larger ZFS pools and workload separation.

RAID Still Does Not Replace Backup

RAID keeps a storage pool available after certain drive failures. Backup lets you recover when files, users, apps, sync jobs, or the whole storage system fail. Those are different problems.

RAID does not protect against accidental deletion, ransomware, fire, theft, a bad update, corrupted sync, failed rebuild, or a user overwriting the wrong folder. ZFS snapshots can help with rollback, but snapshots on the same pool are still not the same as an independent backup.

Protection Layer	What It Helps With	What It Does Not Solve
RAID / RAIDZ / mirror	Drive failure tolerance	Deletion, ransomware, disaster, bad sync
Snapshots	Rollback to earlier file states	Pool loss, theft, fire, all-disk failure
Local backup	Fast restore	Same-site disaster
Offsite backup	Disaster recovery	Fast local restore unless planned well

A safe NAS plan uses RAID for availability, snapshots for rollback, backup for recovery, and offsite copies for disaster protection.

Practical Recommendation by Drive Count

This table is not a law. It is a starting point for home NAS, personal server, and small-office storage decisions.

Drive Count	Practical Choice
2 drives	Mirror
3 drives	Mirror plus backup, or RAIDZ1 only for non-critical data
4 drives	RAIDZ2 / RAID 6 for safety, or two mirrors for performance
5 drives	RAIDZ2 if using HDDs and important data
6 drives	RAIDZ2 / RAID 6 is usually the balanced choice
8+ drives	RAIDZ2, RAIDZ3, or multiple vdev design depending on risk
SSD pool	Mirrors often make sense for performance and latency
Mixed drive sizes	Avoid when possible; plan around the smallest drive

The more important the data, the less you should optimize only for usable capacity. Choose the layout you can rebuild, expand, monitor, and back up without panic.

Final Takeaway

Choose mirrors when performance, simplicity, app data, VM workloads, and easier recovery matter more than maximum capacity. Choose RAID 6 or RAIDZ2 when large HDDs, important data, and multi-drive fault tolerance matter more than saving one drive of space. Use RAID 5 or RAIDZ1 only when the risk is acceptable and backup already exists.

If you are using ZFS, the decision is not only “RAID level.” It is vdev topology. That topology decides how the pool performs, how it fails, how it rebuilds, and how it grows.

The safest RAID choice is not the one with the biggest usable number on day one. It is the one that still makes sense on the day a drive fails.

FAQ

Is RAID 5 still safe for a home NAS?

RAID 5 can still be acceptable for small arrays, SSD pools, non-critical data, or media libraries that already have a full backup. It is not the safest default for large HDD pools or irreplaceable files because it only tolerates one drive failure.

Is RAID 6 better than RAID 5?

For large HDDs and important data, RAID 6 is usually safer because it can tolerate two drive failures. The tradeoff is lower usable capacity and heavier write parity work compared with RAID 5.

Is RAIDZ the same as RAID 5 or RAID 6?

Not exactly. RAIDZ1 and RAIDZ2 are similar to single-parity and dual-parity ideas, but they are part of ZFS vdev topology and work with ZFS features such as checksums, scrub, snapshots, datasets, and send/receive workflows.

Are mirrored drives better than RAIDZ?

They are better for some workloads. Mirrors are often stronger for VM, Docker app data, databases, and random I/O. RAIDZ is often better for large media libraries, archives, backup pools, and capacity-focused storage.

Can I expand RAIDZ later?

RAIDZ expansion has improved in OpenZFS, but it is not a reason to ignore planning. It can add capacity, but it does not automatically change RAIDZ1 into RAIDZ2, remove rebuild risk, or replace backup.

Does RAID replace backup?

No. RAID helps with certain drive failures. Backup protects against deletion, ransomware, sync mistakes, theft, fire, failed rebuilds, and larger storage-system problems. A serious NAS setup needs both RAID planning and backup planning.