NAS, short for network-attached storage, is a dedicated storage server that provides file-level access to data over a standard network, typically through protocols such as SMB and NFS. In server storage, a NAS presents shared folders that many clients can read and write at the same time, while the NAS itself manages the disks, the file system and the permissions.
A NAS can be a compact appliance with two to eight drive bays, a rack-mounted unit in a data center or a general-purpose server running storage software. What defines it is not the hardware but the access model: clients ask for files by name and path, not for raw disk blocks.
How a NAS serves files over the network
A NAS runs an operating system optimized for storage, which combines physical disks into a protected pool using RAID, ZFS or a similar layer, creates a file system on top, and exports folders as network shares. Clients mount these shares and see them as ordinary folders or drive letters.
Because the NAS owns the file system, it handles file locking, permissions and metadata centrally. This is what allows several users to work in the same folder safely, something that raw block storage cannot provide without a cluster file system.
The table below lists the protocols a NAS commonly offers and where each one fits.
| Protocol | Typical clients | Main use |
|---|---|---|
| SMB 3 | Windows, macOS, Linux | Office file shares, user home folders |
| NFS v3 or v4 | Linux, Unix, hypervisors | Server workloads, virtual machine datastores |
| AFP | Older macOS systems | Legacy, replaced by SMB |
| FTP, SFTP, WebDAV | Various | Remote transfer and simple web access |
| S3-compatible object API | Backup and cloud tools | Offered by some NAS systems |
| iSCSI | Servers and hypervisors | Block access, an additional SAN-like function |
NAS compared with SAN, DAS and block storage
The key distinction between storage architectures is the level at which data is shared. A NAS shares files, while a SAN and other block storage share raw volumes that a host formats with its own file system.
| Architecture | Access level | Connection | Typical use |
|---|---|---|---|
| NAS | File | Ethernet, SMB or NFS | Shared folders, backups, media |
| SAN | Block | Fibre Channel, iSCSI or NVMe-oF | Databases, virtualization clusters |
| DAS | Block, single host | SATA, SAS or USB | Local server storage |
| Cloud block storage | Block | Provider network | Virtual machine disks |
| Object storage | Object via HTTP API | Internet or LAN | Archives, backups, large unstructured data |
Many current systems are unified storage: the same device offers SMB and NFS shares and also iSCSI volumes. The protocol chosen for each workload still determines whether the client sees files or a disk.
Performance factors in a NAS
NAS performance depends on the slowest link in the chain between disks and clients. A fast drive pool behind a gigabit link will never exceed about 110-115 MB/s of usable throughput for one stream, roughly the speed of a single modern hard disk.
| Factor | Typical range | Effect |
|---|---|---|
| Network link | 1 GbE to 25 GbE | Caps sequential throughput |
| Drive type | HDD about 75-250 IOPS, SSD tens of thousands | Determines random I/O and IOPS |
| RAID level | RAID 1, 5, 6, 10 or ZFS RAIDZ | Trades capacity, write speed and resilience |
| Memory for caching | 4 GB to hundreds of GB | Larger read cache, important for ZFS |
| SSD cache or special devices | Optional | Accelerates metadata and small random reads |
| Protocol overhead | SMB signing, encryption, sync writes | Reduces throughput and raises latency |
For virtual machines or databases on NFS, synchronous write latency often matters more than bandwidth. Mirrored SSDs or a dedicated log device on ZFS make a larger difference here than a faster network.
Choosing and configuring a NAS
Sizing a NAS starts from capacity, protection level and access pattern, not from drive count alone. Usable capacity is always lower than raw capacity because of parity, spare space and file system overhead.
| Use case | Recommended setup | Reason |
|---|---|---|
| Small office file shares | 4 HDDs in RAID 6 or RAIDZ2, 2.5 GbE | Survives two disk failures at moderate cost |
| Backup target | Large HDDs, snapshots, immutable copies | Cheap capacity and ransomware resistance |
| Video editing | SSD pool or SSD cache, 10 GbE | High sustained throughput for large files |
| Virtual machine storage | Mirrored SSDs, NFS or iSCSI, 10 GbE or faster | Low latency and high IOPS |
| Home media library | 2 HDDs in RAID 1 | Simple redundancy for modest needs |
Good configuration practice includes directory-based permissions tied to central accounts, scheduled snapshots, SMB 3 with signing or encryption where required, disabled legacy protocols such as SMB 1, and email alerts for disk health.
Common NAS misconceptions and risks
The most dangerous misconception is that RAID is a backup. RAID protects against disk failure, but deletion, ransomware, corruption or a failed controller affect every copy at once.
- No off-site copy: a fire, theft or ransomware attack can destroy the NAS and its snapshots together.
- Exposed to the internet: NAS devices published directly through port forwarding are frequent targets of mass attacks.
- Long rebuild times: rebuilding a large HDD in a parity array can take a day or more, during which a second failure is more likely.
- Drive type mismatch: SMR desktop drives can perform very poorly or drop out during rebuilds in some arrays.
- Firmware neglect: appliance updates close vulnerabilities and must be applied regularly.
A sound baseline is the 3-2-1 approach: three copies of data, on two different media, with one copy kept off-site or offline.
A non-technical view
A NAS is a shared filing cabinet connected to the network. Instead of keeping documents on one person’s computer, everyone in a home or office saves them to this central box, and anyone with permission can open them from their own device.
Inside are several hard drives that protect each other, so one broken drive does not lose your files. It still needs a separate backup, just as a filing cabinet does not protect papers from a fire in the same room.
A concrete example
An architecture firm with 15 employees stores project drawings on individual workstations and an external USB drive. Files are duplicated, versions get lost, and nobody is sure where the current plan lives.
The IT contractor installs a six-bay NAS with 12 TB drives in RAID 6, giving roughly 48 TB of usable space before file system overhead, connected over 10 GbE to the core switch. Project folders are shared over SMB 3 with permissions linked to the company directory, and hourly snapshots are kept for two weeks.
A second, smaller NAS at the owner’s home receives a nightly encrypted replica. When a ransomware infection later encrypts one workstation and its mapped project folder, the contractor restores the folder from the snapshot taken an hour earlier. The firm loses less than one hour of work.