TECHNOLOGY
NAS vs Hardware RAID vs Software RAID for a Home Setup
NAS vs Hardware RAID vs Software RAID for a home comes down to whether the disks stay readable after the box or card dies, plus a backup kept off the array.
The photos, the movie files, and the work folder have outgrown the single disk in your PC. The next step is several drives, and NAS vs Hardware RAID vs Software RAID is the comparison that shows up as soon as you shop.
A NAS is a small computer that shares a folder over your home network, while a hardware RAID card groups several drives so one PC sees a single disk. Software RAID is the operating system, or the software inside the NAS, doing that grouping without a special card.
For a household that wants the files on phones and a TV, a NAS running its own RAID software is the fit. One computer can use software RAID in Windows or Linux, including ZFS, and a hardware card fits when that PC should see one disk and you can replace the card with the same family later. At home, the deciding fact is what you can still read after the box or the card stops.
This page covers what each option does, how the common levels split space, what a rebuild reads when a drive dies, which setup matches your files, and the backup that has to sit outside the array.
How a NAS Box Differs From RAID in a PC
A NAS stays powered on so phones, a TV, and laptops can open the same folder, and it runs its own small operating system. Synology builds its SHR layout on Linux RAID, and the other home NAS brands you meet in stores also group the disks in software rather than on a plug-in card you install yourself.
A hardware RAID card sits on the motherboard, or inside a multi-bay shelf, and it groups the drives before the operating system boots so the PC sees one disk. The card can keep the parity math off the computer’s processor, and a model with a battery or flash backup can complete a write after a short power cut. You pay for that card, and a dead card has to be replaced by another from the same card family before those disks will mount again.
Windows, Linux, and systems such as TrueNAS can group the drives in software, with Storage Spaces, mdadm, or ZFS, and they use the processor you already have. On a few hard drives that work stays in the background on a current desktop chip. The disks can move to another computer that runs compatible software, which is the recovery path a locked card does not give you.
| Question | NAS | Hardware RAID | Software RAID |
|---|---|---|---|
| What you are buying | A networked computer that shares files | A controller that builds one volume | Software on a PC or inside the NAS |
| Who can open the files | Devices on your network | The attached PC, unless you share the folder | The computer running the software, unless you share it |
| If the brain dies | Replace the NAS and keep the disks | Replace the card with a compatible one | Move the disks to a machine with the same software |
| Added cost beyond drives | The NAS box | The card | No card on a PC you already own |
| Setup | A wizard in a browser | A card plus a firmware screen | An OS tool, with ZFS asking more than Storage Spaces |
A NAS almost always contains software RAID, so the three labels overlap on a real shelf. You are choosing who manages the disks: a box on the network, a card, or a program. On spinning hard drives, a current PC keeps up with a card for ordinary file sharing. The card earns its place when you want the processor left free, or a write cache that survives a power cut. Put that same card between the CPU and a stack of NVMe drives and the extra hop can hold the drives back, because those drives are already talking straight to the machine.
How the Main RAID Levels Split Your Drives
RAID is the method that makes several physical drives look like one volume, with or without room to lose a disk. Parity is extra math stored across the set so a missing drive can be rebuilt. A mirror is simpler: a full second copy. JBOD, which NAS menus still offer, only stacks the drives, so a dead disk takes the files that lived on it and leaves the other disks alone.
| Level | Minimum drives | Failures survived | Usable space when drives match |
|---|---|---|---|
| RAID 0 | 2 | None | The drives added together |
| RAID 1 | 2 | 1 | One drive |
| RAID 5 | 3 | 1 | All drives but one |
| RAID 6 | 4 | 2 | All drives but two |
| RAID 10 | 4 | One drive in each mirrored pair | Half the set |
Four matching drives make the arithmetic obvious, and the file system still keeps a small slice, so the file manager shows a little less than the table. If the drives do not match, classic RAID clips the set to the smallest disk. RAID 10 does not survive every two-disk failure: lose both members of one pair and that stripe is gone. Stacked layouts such as RAID 50 show up on bigger controllers. A home shelf of four to eight disks does not need them.
Synology SHR on Mixed Drive Sizes
Synology Hybrid RAID is the layout the company points new owners toward, and it is built so mixed disk sizes waste less space than classic RAID on the same box. Synology’s guide to SHR also lists the classic choices still sitting in the menu: Basic, JBOD, RAID 0, RAID 1, RAID 5, RAID 6, and RAID 10.
- SHR-1: Two or more drives, and the pool survives one failure.
- SHR-2: Four or more drives, and the pool survives two failures.
- Model check: Not every Synology chassis offers SHR, so the model page matters before you plan on it.
Stay with SHR when you will buy bigger disks over the years and you expect to keep a Synology box. Pick a classic RAID 1 or RAID 5 on that same NAS when you want a layout a later PC is more likely to understand without Synology’s own layer on top.
OpenZFS RAIDZ and the Write Hole
The OpenZFS description of RAIDZ calls the layout a RAID 5 variation that closes the write hole, the crash case where data and parity no longer match after power loss. One parity disk survives one failure, two survive two, and three survive three. The smallest group is one disk more than the parity count, the suggested width is 3 to 9 disks, and usable space is about the disk size times the number of data disks.
WATCH THE RECORD SIZE
On a three-disk RAIDZ1, the OpenZFS example puts 128 KB records at 66% usable, because 192 KB is written to store 128 KB. The same three disks with 4 KB records fall to 50%, the same ratio as a mirror. Photo and video folders want the larger record. A pile of tiny database pages does not.
Random reads on a general-purpose pool behave like a single disk, so a mirror is the better group when many small files have to come off the disks at once. A movie stream scales with the number of data disks. Checksums let a scrub find a bad block and repair it from a mirror or from parity, which a card that only rebuilds a whole disk never shows you file by file.
Unraid Parity and Mixed Disk Sizes
Unraid keeps each data disk as its own file system and puts parity on a separate disk that has to be at least as large as the biggest data disk. One parity disk can rebuild any single missing data disk, and a second parity disk covers another failure on top of that.
A protected write reads the old data and the old parity, then writes the new data and the new parity. That is four disk operations. Unraid’s own docs list the default write mode at 20 to 40 MB/s. The reconstruct mode, which spins every drive, is listed at 40 to 120 MB/s.
The payoff is the mismatched pile in a closet. Data disks do not have to match each other, as long as parity stays the largest disk in the set. You give up the single-stripe speed of classic RAID 0, and you gain disks you can pull and read as ordinary file systems if the worst day arrives.
Storage Spaces on a Windows PC
The steps and the resiliency types are in Storage Spaces in Windows, and they apply to Windows 10 and Windows 11. You need at least two drives besides the disk where Windows itself is installed. Internal or external SATA and SAS drives are the normal pool. USB can work only when Windows sees each disk as a fixed drive.
- Simple: At least two drives, no protection, meant for scratch such as video render files.
- Two-way mirror: At least two drives, survives one failure, and keeps two copies of each file.
- Three-way mirror: At least five drives, survives two failures.
- Parity: At least three drives to survive one failure, and at least seven drives for protection against two failures.
That seven-drive rule belongs to Storage Spaces, not to classic RAID 6, which starts at four drives. Microsoft aims parity spaces at archives and at music or video you mostly read, and aims mirror spaces at everyday files. Format a mirror with ReFS and Windows maintains integrity checks for you. A pool you upgrade may refuse to mount on an older Windows version, so the move is to another current Windows PC, not to any spare laptop in the house.
Many USB docks never show up in the pool list. Windows rejects disks the enclosure marks as removable, and it rejects docks that hide several drives behind one device. That is how the feature is built, not a driver you can patch around. SATA ports on the motherboard are the clean way to build a Storage Spaces mirror at home.
The Rebuild Read After One Drive Fails
A speed chart never shows the hour after a drive has already failed, when the set is one more failure away from gone. In a four-drive RAID 5 made of 8 TB disks, the repair reads the other three from the first sector to the last. That read is 24 TB.
The maximums in Seagate’s IronWolf datasheet rates are not one number for every model. The table applies each listed maximum to that same 24 TB read, using the decimal terabyte printed on the drive label.
| Maximum nonrecoverable read error rate | Allowance on the 24 TB read |
|---|---|
| 1 per 1015 bits | 0.192 |
| 1 per 1014 bits | 1.92 |
Those figures are ceilings on the sheet, so they choose a layout rather than predicting a given Tuesday. When your drive’s row is the second one, a layout that can lose two disks still has protection after a bad sector on the rebuild. RAID 6, SHR-2, and RAIDZ2 are that layout. A mirror never reads a whole stack of partners to rebuild. It copies the surviving disk.
If the card dies, the disks are still full, and the PC still cannot assemble them until a controller from that same family understands the metadata. A dead NAS mainboard leaves a software array on the disks instead, which is a different repair, and it still wants someone who knows that software. A ZFS resilver walks data the pool marks as out of date after you replace a disk, and a later scrub checks checksums across what remains.
The time for a classic parity rebuild is however long the slowest surviving drive takes to be read completely, and the volume is slower for everyone while that read runs. Copy the files you cannot replace before you start, while they are still visible. The rebuild is the stressful window, not the moment the first drive’s light went amber.
Picking a Home Storage Setup for Your Files
Once you know what that rebuild reads, the setup is about who opens the files and how painful a dead disk would be.
MATCH THE SETUP TO THE JOB
- Family files: Use a two-bay NAS in RAID 1 or SHR-1 for photos and documents. Move to four bays with RAID 6 or SHR-2 when the drives are large and this box holds the only originals for a while.
- One editing PC: Use software RAID on that computer. Mirror the files you cannot replace, and use a simple space or RAID 0 only for scratch that already has a copy somewhere else.
- Mismatched disks: Use Unraid, or SHR if the box is a Synology. Classic RAID will clip every disk to the smallest member.
- A card you own: Use it when one PC needs a single protected volume. The replacement card has to be the same family, so the model number belongs on a sticker.
- PC stays off: Use a NAS. A RAID volume inside a desktop disappears from the house the moment that PC sleeps.
Remote access follows the NAS, because the box has its own system and can open a path from outside if you turn that on and keep the password strong. A RAID volume in a PC is local until you add file sharing, and that share dies with the PC. None of the rows above replaces a copy that is not a member of the set.
What to Do When One Drive Drops Out
The level you picked only helps if the first failure is handled in order. The volume can still be online and already one disk away from a total loss.
DO NOT CREATE A NEW ARRAY
Initializing the set, building a fresh array on the same disks, or forcing a repair overwrites the membership records a later recovery would need. If the files are irreplaceable and they no longer open, stop and image the members instead of retrying the rebuild that just failed.
- Leave the array alone: Do not initialize it, do not create a new one on those disks, and do not run a disk repair that writes to a volume which just vanished.
- Copy first: If the files are still visible, copy the ones you cannot replace onto a drive that is not a member of the array.
- Label the bays: Photograph the slots before you pull anything, and power the box down cleanly first when the manual says to.
- Check the survivors: Pending sectors on a second disk are a reason to keep copying, not a reason to start the rebuild.
- Then replace: Use a drive of equal or larger size, and start the repair from the NAS page, from Storage Spaces, or from the ZFS replace action. On a hardware card, use that card’s own utility.
A hot-swap bay lets you pull the failed disk while the box stays on, which is the easy part of a hardware shelf and of many NAS units. The hard part is the read that follows. If a second disk drops during that read, the single-parity set is finished, and the copy you made in step two is the copy you still have.
A Separate Backup Beside the Home Array
A RAID set keeps the folder online after a disk dies. A deleted folder, ransomware encryption, or a fire in the room is written straight through a mirror, and parity has no older version of that file unless you also kept snapshots.
The home version of CISA’s 3-2-1 backup rule is three copies of the important files, on two kinds of storage, with one copy kept away from the house. On a real shelf that is the array, a disk that is not a member of the array, and either a cloud backup or a drive you store at another address.
Snapshots on the NAS save you from a bad delete, and they die with the disks if that is the only place they live. An unplugged drive in a drawer covers a dead box. The copy outside the house covers the house itself. Restore a few files on purpose before you trust the job, because a backup you have never opened is still a guess.
Common Mistakes With NAS and RAID at Home
Most lost home arrays come from a step taken after the first disk failed, or from a layout that never had a spare copy in the first place.
MISTAKES THAT COST THE FILES
- RAID 0 photos: A stripe has no spare, so one dead drive takes the whole set, including the only copy of the pictures.
- Rebuild first: The repair reads every surviving disk, so copy the unique files off while the volume is still up.
- Chipset menu: A RAID screen in a desktop BIOS still leans on a driver. It is not a standalone card, and it fights ZFS or Storage Spaces if you leave it on.
- SMR parity: Recovery labs see drive-managed SMR disks stall long enough for a controller to drop them. Prefer drives sold for NAS that advertise error recovery control.
- One aging set: Disks bought on the same day can wear on a similar schedule. On large drives, give the set two-failure protection or keep the backup current before a long rebuild.
- No sticker: The card model, the RAID level, and the software name are what a later repair needs, and a label inside the case holds them.
- Open remote access: A NAS share pointed at the internet with a weak admin password is a network door, not a private disk.
A power cut during a parity write can leave data and parity out of agreement. A hardware card with a battery-backed cache is built to finish that write, and ZFS avoids the mismatch by the way it lays the stripe down. A desktop software array with neither protection is a reason to put the box on a small UPS when these files matter.
Frequently Asked Questions
Can You Run Software RAID Without a BIOS RAID Menu?
Yes. Software RAID only needs the motherboard to present each drive to the operating system, so a board with no RAID option still works with Storage Spaces, mdadm, or ZFS. A RAID line in the BIOS is a separate feature, and turning it on can hide the drives those tools need to see.
How Often Should You Scrub a Home ZFS Pool?
A monthly scrub is the home baseline in the OpenZFS guidance, and it is how a bad block gets found while a mirror or parity can still repair it. Start the scrub when a resilver is not already running, because ZFS runs only one of those two jobs at a time.
What Does a Foreign Configuration Warning Mean?
The card sees array metadata it did not create under its current settings, which shows up after a card swap or a battery problem. Do not import it, initialize the disks, or build a new array until you know the prompt belongs to the original card, because the wrong yes can erase the map to your files.
Does a One-Drive SHR Volume Have Any Protection?
No. With only one drive in the pool, SHR has nothing to rebuild from, and Synology’s RAID table treats that volume as having no failure tolerance. Add another drive before you count the pool as protected.
Will a Desktop Hard Drive Work in a Home RAID Set?
It can, and a data-recovery note is worth reading before you buy. Those write-ups describe NAS drives as reporting a bad sector in about 7 seconds, while a desktop drive can keep retrying for 30 seconds to 2 minutes. Controllers in the same notes give up after 8 to 20 seconds, so a desktop drive can be marked failed while it is still trying, and the rebuild gets harder. For a parity set, pick drives that advertise error recovery control.
Put a sticker in the case with the card model or the software name, and keep one copy of the files on a disk that is not a member of the array. Store that copy outside the house. The layout you picked will handle a dead drive. The sticker and the outside copy are what handle a dead box.
-
NEWS4 months agoMicrosoft’s 96% Cyber Score Sits at 86.3% on the Board
-
BUSINESS6 months agoMusk’s $914 Billion Lead Is a Public SpaceX Bet
-
ENTERTAINMENT1 year agoSkrilla’s 6-7 Became Word of the Year by Meaning Nothing
-
SPORTS4 months agoFree Live Sports Streaming Without Cable: An Antenna and App Guide
-
GAMING2 months agoXbox Puts 15-Hour Caps on Game Pass Cloud Play
-
ENTERTAINMENT4 months agoThe Odyssey’s IMAX Film Run Hit a 41-Theater Limit
-
ENTERTAINMENT5 months agoEndgame Encore’s $86 Million Trial for Infinity Vision
-
BUSINESS2 years agoRetail Banking vs Commercial Banking: Who Each Side Serves
