Building a NAS can seem deceptively simple. After all, at its foundation, it's a computer with a hunk of storage, running some kind of operating system that allows you to access its storage volumes from your network. Like many things in the self-hosting world, however, it's rarely that straightforward. The problem is that most of the decisions that matter don’t let their ramifications be known until months or even years later. By then, changing course is expensive, disruptive, or downright impossible without migrating all your data. Looking back, there are a few lessons I wish someone had drilled into me before I ever laid my hands on a drive bay.

Plan your storage beyond just capacity

The "how" matters more than "how much"

One of the biggest mistakes one can make when building a NAS isn't buying too little storage, but instead only planning for how much you need today.

You might not be able to afford 4 of the same high-capacity NAS drive today, so you settle on 2 drives, leave 2 bays open, and call it a day. This will work for a while, until you need to expand, and that often means replacing drives one by one, waiting for long rebuilds, or discovering that your chosen redundancy scheme wastes far more capacity than you expected once you start mixing drive sizes.

Planning beyond raw capacity means thinking about how you’ll add storage in two or three years, not just how much you need this month. It means choosing drive sizes that leave room to grow, understanding how your array handles mismatched disks, and accepting that buying slightly more storage up front can not just save you money, but save you the stress of having to reconfigure everything down the line.

Redundancy isn't the same as a backup

This is probably the most important lesson about building a NAS

This is unfortunately the lesson that a lot of users learn the hard way. Redundancy keeps your NAS online when a drive fails, but it does nothing to protect your data from most real-world disasters. Accidental deletions, overwritten files, ransomware, software bugs, filesystem mishaps, or even a catastrophic power event can wipe out perfectly redundant arrays in seconds.

RAID is not a backup. It buys you time to replace a failing disk, and not much else. Also, once something bad is written to the array, redundancy happily mirrors or parity-protects that mistake across all drives.

A real backup lives somewhere else completely, ideally off the NAS entirely. That could be another machine, an external drive rotated periodically, or even cloud storage. That last point might feel like it defeats the purpose, but the key point is separation. If your NAS dies, your backup should still be able to happily exist elsewhere.

Buy used enterprise HDDs

It's the way to go

Buying brand-new consumer hard drives feels like the safe choice, but for a NAS, used enterprise drives are often the smarter one. Decommissioned datacenter drives are built for continuous operation, vibration-heavy environments, and predictable workloads. They're designed to be run hard and put away wet, and that's perfect for a consumer NAS.

They're usually much larger than consumer drives, and are competitive on cost-per-terabyte. In a lot of cases, these used drives aren't decommissioned due to a failure, but rather because of a fixed replacement schedule, which is done to reduce risk of data loss as much as possible. Even if they are drives that previously failed, there are many storefronts that take them, refurbish them to a high standard, and then sell them at a discount. Failure rates of these used enterprise drives follow a "bathtub curve", meaning they either fail extremely early in their lifespan, or extremely late, and as long as you double-check the SMART history, you should have no problems happily driving used enterprise HDDs for years.

Networking is usually the main bottleneck

SATA drives are not going to be the source of slowness

It's actually very easy to build a NAS that's able to put out more bandwidth than what your network can handle. Modern HDDs, efficient filesystems and an SSD cache (if you're feeling fancy) can easily saturate a 1 GbE connection.

Not everyone needs a multigig connection, especially if what you're doing won't exceed 125 MB/s, which is often the upper limit of a 1 GbE link in ideal conditions, but it is worth looking at if speed is a priority. I wouldn't fill up your NAS with Gen 5 NVMe drives without at least having 2.5 GbE, if not more.

A NAS can be an old PC, but there's more to consider

You can easily turn an old PC into a NAS, but it's important to consider more than just storage capacity and operating system when doing so. You can save a lot of future headaches by planning your storage scheme well, making backups, buying used enterprise drives, and setting proper expectations of your transfer speeds.