How to choose RAID for NAS? I've used RAID1, RAID5, and RAID6 all. At home, just use a few drives to match your numbers and don't copy your homework

the most confusing question on the day of installing a NAS is this question. Feiniu's storage wizard has a row of options: Basic, RAID1, RAID5, RAID6. Some in the group call out "brainless RAID5," while others advise, "For home setup, what array? Single drive plus backup is the right way." My first machine was randomly RAID1, then after getting four bays, I built RAID5. In between, I even paired RAID6 for a friend's six-bay machine, exploring all three types and writing down the real accounts and pitfalls in this article. Let me start with the harsh truth: RAID doesn't have the best option, only matches your number of plays and usage.

first explain what RAID does: it handles bad disks, not the wrong deletions

many people think that building RAID data means everything is fine, but this is the biggest misconception. RAID stands for Independent Disk Redundancy Array. Simply put, it has several disks copy each other's jobs: if one disk is broken, the remaining disk still retains complete or calculable data, so the machine doesn't shut down and files aren't lost. It deals with the kind of death where the hard drive itself is broken.

listing things it can't control, you'll feel a chill in your heart: files deleted by hand, ransomware encryption, the whole machine being struck by lightning, water damage, and you can't move it. In these four situations, all disks suffer simultaneously, and the homework copied from each other disappears together. So I'll set the rule here: RAID is availability, not backup. The truly stable approach is to make a backup outside the array. I covered the 3-2-1 approach in the of the backup, but today I'll focus on how to choose the array itself.

four commonly used modes for breakdown: available capacity and bad disk tolerance

directly show my own and the real configurations I helped my friend make—just do the math and you'll understand.

Basic (no array) : Several disks each operate independently; three 8T blocks make three independent 8T spaces. The advantage is that no storage is wasted; you lose data wherever it fails, and other drives don't have to worry about it. My mom's single-drive N100 for running Flying Cow for photo storage is basically Basic, because I backed it up separately, so the array was actually redundant.

RAID1 (Mirror) : Two drives are backups for each other, two 4TB drives still have 4TB available capacity, and half capacity is insurance. This is how I built my first Synology DS220+. It has a hidden benefit that many people don't know: if one drive breaks, you can detach the other one separately and plug it into any computer to read all files directly, because essentially, it's a complete disk. This trait is valuable when saving data.

RAID5 (Single Validation) : At least 3 disks, with usable capacity minus one tray. Currently, I have four 8TB RAID5 blocks and 21.7TB available, with each block accounting for about one-third of the redundant overhead. If any piece is damaged, the remaining disk can recalculate the data; during the bad drive, the second chip is damaged before it's finished.

RAID6 (Double Verification) : At least 4 disks, capacity minus 2 disks; if two disks fail simultaneously, it can withstand it. My friend's six-bay machine is exactly this one for him. The six-yuan 16TB can be used for 64TB. Why he went to RAID6 in detail later is directly related to the rebuild time.

Don't just look at capacity accounts: rebuilding time is the heaviest hidden cost for household use

this is the point I truly understood after suffering losses. A RAID bad drive isn't just about replacing it with a new drive; it recalculates the data and writes it back into the new drive. This process is called rebuild. During rebuild, the array is in a 'downgrade' state, with no redundancy left. If another piece breaks at this point, the entire pool's data will be wiped out.

reconstruction speed is directly linked to capacity. I built four 8TB RAID5 drives. During the rebuild of a broken drive, it took about eleven hours, with an average write speed of around 180MB/s, and that's still when both the drive and the machine are healthy. My friend's six-yuan 16T car spent twenty-six hours rebuilding to replace the discs—over a whole day. During this time, his whole family had to be extremely careful with the records, basically running naked for a whole day and a half. What if the market is even bigger? Now, for disks over 20T, it's easy to rebuild for two or three days. The larger the drive, the longer the rebuild window, and the higher the probability of secondary failures in the window. This is why RAID5 is starting to be unreliable in the era of large capacity.

so my logic for putting a friend in RAID6 was clear: he had six 16T units, and rebuilding took over a day. If the old property next to him couldn't hold up halfway through (bought the same year, same age), RAID5 would be cremated immediately. RAID6 spends two extra disks of space, but what you gain is the confidence that another drive can break during rebuild. If the data isn't worth the price, of course you don't need to go for it. His family has all the photos and videos of their two children from birth to now, and he thinks it's worth it.

just a few plates and stick to the situation—don't copy others' homework

the number of discs is set, but there aren't that many branches to choose from. I'll just give a conclusion based on the number of discs:

two discs: Don't even think about it, RAID 1. Dual-bay machines have no second option, so trading capacity for security is widely recognized as a cost-effective deal. If someone says "split into two separate drives to double the capacity," you can try, but to be clear about whether you accept losing half the data if one piece breaks, I don't recommend home users take a tightrope.

three drives: RAID5 is the mainstream solution. Three 8TB drives can use 16TB, which costs one-third. Some even build three RAID1 drives (still one block of data is available), purely because the data is extremely valuable and backup obsessives are unnecessary. Single drive or three Basic blocks are only recommended for those who have a separate backup for each block.

Four Plates: The watershed is right here. For medium and small capacity drives under 8TB, RAID5 rebuild lasts about ten hours, with controllable risk windows and a 75% capacity utilization rate, which looks good. My four 8TB drives are RAID5. But if you go for a large 16T or 20T drive, rebuilding can take a day, so I prefer RAID 6: four 16T blocks can be used for 32T, with capacity halved but a good sleep. Some people build two RAID1 sets of four drives. The advantage is that the two sets are unrelated—a broken block only affects the other, but the downside is only half the capacity remains. The cost is the same as RAID6, but you can only afford "one broken piece per group." It's better to just use RAID6 for convenience.

Five Blocks or higher: The more disks there are and the larger the single disk, the more recommended it is to start with RAID6. At the eight-game and ten-game game of the MindTech game, Synology still has more lively gameplay like SHR, Unraid, and ZFS, which could be dedicated in a separate article, but I won't elaborate here.

the three pitfalls I've stepped into, taking a quick look before forming the array would save you trouble

First pitfall: Buy all models at once, but don't buy the same batch. RAID requires the same disk capacity. If 4TB and 8TB are mixed, the array will focus on the smallest portion and waste the larger portion. My first machine was a 4TB one, and after half a year I wanted to add another one. The same model went up by 60 yuan, so I bought another one with the same capacity from another brand. Setting up RAID 1 works fine without any issues. But don't buy disks from the same batch on the same day—a batch may have the same factory defect, causing both sides to fail and rendering the array meaningless. Buying from different brands at different times is actually a kind of outdated safeguard.

second pitfall: think carefully before forming an array, but you basically won't be able to change it later. When creating RAID mode, the drive is initialized—switching from RAID1 to RAID5, expanding from RAID5—all require data to be exported and redone (some systems support on-site migration, so home cables from Feiniu and Synology are usually a hassle). So before buying, first think through positions and capacity planning get everything done in one go. Don't flip through data twice like me, starting with two yuan and then four units.

Third pitfall: Don't panic about bad stocks, but control yourself during the downgrade. If you notice an array alarm, first check which drive is broken and check the SMART health status, then quickly buy a drive with the same capacity to fix it (no need for the same brand, just enough capacity). During the downgrade, avoid writing large files and avoid intensive read/write outside of the array scrub, allowing the machine to quietly complete the rebuild. During the reconstruction of my house, I only went to sleep after running with it in my ears. Although I rationally knew the chances were slim and emotionally uncomfortable,

Conclusion: Just follow the order of decisions

first give the order: First, is the data backed up elsewhere? Yes, Basic works too. No, keep going. 2. How many plates? Two RAID1 drives, three RAID5 drives, four disks depending on disk size—medium and small RAID5, 16TB up RAID6, and more than five disks for mindless RAID6. 3. Rebuild time calculates yourself: available capacity divided by 150MB/s, if it exceeds 20 hours, seriously consider double verification.

the last two honest words. First, RAID buys the device for non-downtime when hardware fails, not a data safe. Important photos need to be backed up—two separate matters. Second, don't scrape together the array just for the sake of it—I've seen people buy two extra trays just to build RAID5, but most of the capacity is idle, so the array needs to be a concern. Let the plates you have run first, add more when it's insufficient, and when adding them, just follow this set of assigned assignments. As for how to configure the write cache after the array starts running, and whether the SSD adds acceleration, that's another matter. I've calculated it in this solid-state cache test .