NAS (Network Attached Storage)

A NAS is, at its core, a small dedicated computer with multiple drive bays, a stripped-down Linux operating system, and just enough networking software to act as a file server. The term has been around since the late 1990s, but consumer adoption took off in the late 2000s when Synology and QNAP made user-friendly NAS units cheap enough for home users and small businesses. By 2026, the global NAS market is estimated at $40 billion, growing at roughly 15% per year as the volume of personal media and SMB data outpaces what people want to store in the cloud.¹

The hardware inside a NAS

Open any consumer NAS and you’ll find a roughly similar set of components. The cheapest 2-bay home unit and the most expensive 24-bay rackmount unit differ in scale, not in architecture:

• CPU. Usually an ARM processor on entry-level units (Realtek, Annapurna Labs) or an Intel/AMD x86 processor on mid-range and higher (Celeron, Atom, Ryzen). The CPU runs the NAS operating system, handles file protocol translation, and performs RAID parity calculations on software-RAID models. More powerful CPUs enable features like real-time video transcoding and on-the-fly encryption.
• RAM. 1 to 32 GB depending on the model. RAM matters for caching file metadata, running ZFS (which is RAM-hungry), and supporting concurrent user sessions. Most home units ship with 1-4 GB; SMB units with 4-16 GB; enterprise units with 32+ GB.
• Drive bays. Hot-swappable bays for 2.5-inch or 3.5-inch hard disk drives, or for SATA SSDs. Modern mid-range units also include M.2 NVMe slots for SSD caching. Bay counts range from 1 (rare) to 24+ in rackmount units.
• Network interface. 1 GbE Ethernet on entry-level units, 2.5 GbE or 10 GbE on mid-range and higher. Most consumer NAS units have at least two Ethernet ports for link aggregation or failover.
• Boot device. Usually an internal flash module (called a DOM, “Disk-On-Module”) or a tiny eMMC chip that holds the NAS operating system. The data drives never hold the OS; if the boot device fails, the data drives are usually still readable.
• Optional NVMe cache. M.2 SSD slots that can be used as read cache, read/write cache, or as a separate fast volume. Critical for performance on metadata-heavy workloads but introduces additional failure modes when configured as write cache.²

The NAS operating system

Every consumer and SMB NAS runs a custom Linux-based operating system with a web management interface. The OS handles RAID, file systems, network protocols, user permissions, snapshots, and increasingly a sprawl of additional applications (media servers, photo backup, container hosting, virtual machines). The major NAS operating systems and their characteristics:³

• Synology DSM (DiskStation Manager). The most polished consumer NAS OS, with strong app ecosystem, excellent backup tools (Hyper Backup, Snapshot Replication), and proprietary Synology Hybrid RAID (SHR) that allows mixing drive sizes. File system: Btrfs by default on modern units, ext4 on older units.
• QNAP QTS. Linux-based with ext4 file system on top of standard Linux md-RAID. More raw hardware features at similar prices (more CPU power, hardware video transcoding) but with a steeper learning curve and more complex application catalog.
• QNAP QuTS hero. A separate QNAP operating system based on ZFS instead of ext4. Offers ZFS features (snapshots, replication, copy-on-write, checksumming) but uses different RAID and storage architecture from QTS, requiring different recovery techniques.
• WD My Cloud OS. A simpler Linux build with ext4 file system. Lower price, lower complexity, and a more limited app ecosystem.
• NETGEAR ReadyNAS OS. Linux with Btrfs. Less common in 2026 but still maintained for existing units.
• TrueNAS / FreeNAS (open-source). Community-developed NAS OS based on FreeBSD or Linux with ZFS. Used by enthusiasts who build their own NAS hardware.

File sharing protocols: SMB, NFS, AFP

A NAS exposes its storage to the network through file-sharing protocols. Different operating systems prefer different protocols, and most NAS units speak all of them simultaneously:

• SMB (Server Message Block). Windows’ native protocol, also widely used by macOS and Linux. The dominant file-sharing protocol on consumer and SMB networks. SMB 3.x adds encryption, multichannel, and significant performance improvements over older versions.
• NFS (Network File System). Unix and Linux’s native protocol. Often faster than SMB for Linux clients, and the standard choice for VM hosts that store virtual disks on a NAS. NFSv4 adds Kerberos authentication and stateful operations.
• AFP (Apple Filing Protocol). Apple’s legacy protocol, deprecated in macOS 11 in favor of SMB. Most modern NAS units still support AFP for backward compatibility with older Macs and Time Machine backups, but new deployments should use SMB.
• Other protocols. WebDAV, FTP, iSCSI (block-level instead of file-level), and S3-compatible API are commonly available on mid-range and higher NAS units.

RAID inside a NAS: the abstraction stack

Almost every multi-bay NAS uses RAID internally to provide redundancy across multiple drives. What’s important for recovery is that NAS storage is not a single layer; it’s a stack of three abstractions, each of which can fail independently:

1. RAID layer (md-RAID or hardware RAID). Combines multiple drives into a single logical volume. Most consumer NAS units use Linux mdadm software RAID. The RAID layer presents a single block device to the next layer above.
2. LVM layer (Logical Volume Manager). Optional but common on mid-range and higher units. Splits the RAID block device into one or more logical volumes, supports thin provisioning, snapshots, and dynamic resizing. Adds another layer of complexity to recovery.
3. File system layer. Where files actually live. ext4 on most QNAP and WD units, Btrfs on Synology, ZFS on QuTS hero and TrueNAS. Each file system has its own corruption modes, repair tools, and recovery characteristics.

Synology adds a fourth layer: SHR (Synology Hybrid RAID), which is a custom layout on top of md-RAID that allows mixing drive sizes. QNAP’s QuTS hero uses ZFS that handles RAID and the file system together with no separate LVM layer. The recovery approach differs for each combination, which is why NAS recovery is brand-specific in practice even though the underlying technologies are mostly Linux-standard.⁴

Three networked storage architectures get compared constantly because they sound similar and serve overlapping purposes. They’re actually different at a fundamental level: where the file system lives, what the network sees, and what kind of workload each one is best for.⁵

DAS (Direct Attached Storage)

Storage that connects directly to one computer through USB, Thunderbolt, eSATA, or a SAS/SATA cable inside a server chassis. The simplest case: an external hard drive plugged into your laptop. DAS is fast (no network overhead), cheap, and simple, but only the host computer can access the data. Use cases: working storage, backups, single-user setups. DAS predates NAS and remains the default for individual users.

NAS (Network Attached Storage)

A file-level storage server on the network. Multiple clients access shared folders through SMB, NFS, or AFP. The NAS handles the file system internally; clients see folders and files. Strengths: easy setup, multi-user access, RAID redundancy, lots of vendor options at consumer prices. Weaknesses: limited performance for high-IOPS workloads (databases, virtualization), bottlenecked by the network speed.

SAN (Storage Area Network)

A block-level storage network. The SAN exposes raw disk volumes to servers over a high-speed network (Fibre Channel, iSCSI, FCoE), and each server’s operating system handles the file system on the volume it’s been assigned. SAN is the architecture for enterprise databases, large-scale virtualization, and any workload where storage performance is the bottleneck. Strengths: maximum performance, can support thousands of concurrent connections, advanced features like LUN masking and zoning. Weaknesses: expensive, complex to design and manage, requires specialized network hardware.

Three architectures at a glance

Property	DAS	NAS	SAN
Storage level	Block (single host)	File	Block (multiple hosts)
Connection	USB, Thunderbolt, SATA	Ethernet (LAN)	Fibre Channel, iSCSI
File system location	Host computer	NAS device	Each connected server
Multi-user access	No (one host)	Yes	Yes
Typical use case	Personal storage, backup	File sharing, SMB	Database, virtualization
Cost	Lowest	Medium	Highest
Setup complexity	Plug and play	Web UI configuration	SAN admin required
Performance	High (local bus)	Network-limited	Highest (dedicated network)

In 2026, the lines blur somewhat. Higher-end NAS units (Synology XS+, QNAP TVS-h series) support iSCSI for block-level access alongside file-level access, effectively functioning as small SANs. Hyperconverged infrastructure further blurs the distinction by combining compute and storage on the same physical hardware. For most home and SMB use cases, NAS is the right choice; SAN only makes sense for large-scale enterprise deployments.

NAS failures combine the failure modes of individual hard drives with the failure modes of RAID arrays, plus several additional failure modes specific to the NAS appliance itself. Most are recoverable, but the recovery cost varies enormously based on what failed and what the user did before calling for help.⁶

• Single drive failure on a redundant volume. The expected scenario. RAID 1, 5, 6, SHR, and similar redundant configurations continue running on the surviving drives. The NAS sends an email alert (if configured) and shows the volume as “degraded.” Replace the drive, let the NAS rebuild, done. No recovery needed.
• Multiple drive failures or rebuild cascade. The catastrophic scenario covered in detail on the RAID page. Two failed drives on RAID 5, three on RAID 6, both drives in a mirror pair on RAID 10. Recovery requires lab-level imaging and virtual array reconstruction. Cost runs $1,500 to $5,000 for typical 2-4 bay home and SMB units.
• “Volume Crashed” / “Storage Pool Inactive” errors. Synology’s “Volume Crashed” and QNAP’s “Storage Pool Inactive” are status messages that can originate at any of the three abstraction layers (md-RAID, LVM, file system). The fix for each layer is different. Vendor support typically suggests destructive options like reinitializing the volume; recovery labs work nondestructively from cloned images.⁷
• ext4 journal corruption. The most common file system failure on QNAP and WD units. Triggered by sudden power loss, kernel panic, or interrupted firmware updates. Symptoms: volume mounts as read-only, missing folders, I/O errors on specific files. Recovery is journal replay and inode tree reconstruction on cloned images, recoverable by labs in most cases.
• Btrfs corruption (Synology). Synology’s default file system has rare but distinctive corruption modes: checksum mismatches that mark files as unreadable, snapshot tree corruption, and orphaned subvolumes. Recovery requires Btrfs-aware tools and can sometimes be performed in-place using btrfs-check, but extensive corruption needs full lab reconstruction.
• NVMe SSD cache failure. When NVMe SSDs configured as write cache fail before flushing dirty data to the HDD pool, the file system on the HDD pool is left inconsistent. Symptoms: files appear corrupted on the HDD volume, recent writes are missing entirely. Recovery requires imaging the cache SSDs and reconstructing the pending writes back to the pool.⁸
• Failed firmware update. A power loss or network interruption during a NAS firmware update can corrupt the boot device (DOM or eMMC). The NAS won’t boot, but the data drives are usually unaffected. Vendor recovery tools (QNAP’s manual NAS Recovery Guide, Synology’s recovery mode) can sometimes restore the boot firmware. If those fail, the data drives can be imaged and reconstructed externally.
• Ransomware attacks. Internet-exposed NAS units have been targeted by major ransomware campaigns: Qlocker (2021), DeadBolt (2022), and ChernoLocker (2023) hit QNAP units; SHIeldNAS hit Synology in 2023. The attackers exploit unpatched vulnerabilities in vendor firmware to encrypt all files on the volume. Recovery depends on having pre-attack backups; paying the ransom is rarely effective.⁹
• Power surge or simultaneous power loss. A power event can damage the NAS appliance, the drives, or both. Without a UPS, sudden power loss during writes commonly leaves the file system in an inconsistent state that journal replay can’t fully fix. Power surges through Ethernet (lightning strikes) can destroy the NAS entirely while the drives survive intact.
• Accidental volume deletion or reinitialization. The single most preventable disaster, and unfortunately one of the most common. A user clicks the wrong button in the web UI; the NAS dutifully reinitializes the volume, wiping all data. Recovery is possible if the underlying data hasn’t been overwritten, but every minute the NAS continues to operate after the deletion overwrites more recoverable data.

Vendor-specific recovery considerations

The recovery approach differs by vendor because each vendor uses different RAID layouts, LVM configurations, and file systems:

• Synology. Uses Linux md-RAID with proprietary Synology Hybrid RAID (SHR) extensions that allow mixing drive sizes. Default file system is Btrfs (modern units) or ext4 (older units). Drives can usually be moved to another Synology unit of similar generation but not to other vendors.
• QNAP QTS. Linux md-RAID with ext4 file system, often with LVM for thin provisioning. Standard Linux storage stack, well-understood by recovery labs. Drives can be moved between QNAP units of similar generation.
• QNAP QuTS hero. ZFS-based, with completely different recovery techniques than QTS. ZFS’s checksumming and copy-on-write features can sometimes recover from corruption that would destroy ext4, but ZFS pool damage is harder to repair than ext4 journal corruption.
• WD My Cloud. Custom Linux with ext4. Limited migration options; drives are typically not portable to non-WD hardware.
• NETGEAR ReadyNAS. Btrfs with custom RAID layout. Recovery requires labs familiar with the specific ReadyNAS firmware version.
• Buffalo TeraStation / LinkStation. Older units use ext3/ext4 with custom XFS variants on newer enterprise models. Very vendor-specific recovery; few labs handle Buffalo well.

Warning signs your NAS is failing

NAS units are designed to surface warnings clearly, but only if email alerts are configured (most users skip this step during setup). Watch for these symptoms in combination:

• SMART warnings on individual drives in the storage manager. Replace drives showing reallocated sectors, pending sectors, or rising read error rates before they fail outright.
• Drives marked as “warning” or “abnormal” in DSM Storage Manager (Synology) or Storage & Snapshots (QNAP). This is the NAS telling you a drive is on its way out.
• Volume status changing from “healthy” to “degraded” usually means a single drive failure on a redundant volume. The NAS is still operational but has zero remaining fault tolerance until you replace the drive.
• Sudden drop in performance, especially on writes. Could indicate bad sectors on a member drive, full or failing NVMe write cache, or LVM thin provisioning running out of physical space.
• NAS unresponsive to web UI but file shares still working. Often a sign of the management interface crashing while the underlying file services are still running. A graceful shutdown via SSH is safer than power-cycling.
• Beeping from the NAS. Synology and QNAP both beep on critical events: fan failure, drive failure, temperature warnings. Check the email alerts (if configured) or the web UI to identify the specific issue.
• Inability to access shared folders after a firmware update. Could be a botched update (rare with major vendors), a permission misconfiguration, or in worst case, file system corruption that the update triggered.
• Repeated drive disconnections. Often caused by a marginal SATA backplane connector, vibration in the chassis, or a failing power supply. Open the NAS and reseat drives if you’re comfortable; otherwise let the vendor or a recovery lab handle it.

Consumer and SMB NAS units come in three rough categories based on physical form factor and use case. The choice affects price, expansion options, and noise level.

Category	Bays	Capacity (2026)	Typical price	Best for
1-bay home	1	4–28 TB	$150–$300	Single-user backup, no redundancy
2-bay home	2	8–56 TB	$200–$500	Home backup with RAID 1, photo libraries
4-bay SMB	4	16–112 TB	$500–$1,500	Small office, RAID 5/6/SHR, media server
6-8 bay SMB	6–8	24–224 TB	$1,000–$3,500	Larger SMB, video editing, virtualization
Rackmount enterprise	12–24+	56–672 TB	$3,000–$30,000+	Enterprise file server, scale-out

1-bay units are essentially network-accessible external hard drives. They have no redundancy and offer little advantage over a USB drive shared from a desktop. Useful only when network access is the primary requirement. The 2026 capacity ceiling is 28 TB single-drive Seagate Expansion HAMR drives.

2-bay units are the dominant home NAS form factor. Two drives configured as RAID 1 give you a single drive’s worth of capacity with redundancy. Synology DS220+/DS224+, QNAP TS-264, and WD My Cloud EX2 are typical. The 2026 entry-level units start around $200 with no drives included.

4-bay units are the most popular small business and prosumer choice. Four drives in RAID 5, RAID 6, or SHR provide a good balance of capacity, redundancy, and rebuild risk. Synology DS923+/DS424+, QNAP TS-464, NETGEAR RN424. Most include M.2 NVMe slots for SSD caching.

6-8 bay units serve larger SMBs and prosumer use cases like video production. RAID 6 becomes the standard configuration at this size because rebuild times on RAID 5 with 6+ large drives become impractical. Synology DS1823xs+, QNAP TS-873A. Often include 10 GbE networking and more powerful CPUs.

Rackmount enterprise units are dedicated server-room hardware: 1U, 2U, or 4U rack form factors with hot-swap drive bays, redundant power supplies, and 10 GbE or 25 GbE networking. Synology RS series, QNAP TS-h series, NetApp, Pure Storage, EMC. These overlap with traditional enterprise SAN territory and are usually managed by IT departments rather than individual users.¹⁰

NAS solves a specific set of problems well: centralized storage with multi-device access, drive-failure redundancy, and a manageable UI for non-IT users. It’s not a universal solution; it’s a specific point on the cost/complexity/capability curve.

NAS vs other storage approaches

Property	NAS	External HDD	Cloud Storage	SAN (enterprise)
Multi-device access	Yes	No	Yes	Yes
RAID redundancy	Yes	No	Provider-dependent	Yes
Capacity (typical)	4–200 TB	up to 28 TB single drive	Effectively unlimited	Petabytes
Cost (typical)	$200–$3,000 + drives	$80–$500	$5–$30/TB/year	$10,000+
Speed (typical)	100–1,000 MB/s (LAN)	100–1,000 MB/s	Limited by internet	500–10,000 MB/s
Setup complexity	Web UI wizard	Plug and play	Account creation	SAN admin required
Off-site backup	Manual / app-based	Manual	Built-in	Replication features
Vendor lock-in	High (vendor-specific)	Low	Medium-high	High
Best for	Home/SMB shared storage	Single-user backup	Off-site backup, sync	Enterprise databases

NAS advantages and drawbacks