NAS & RAID Glossary

Quick definitions of all the important NAS, RAID and storage terms. Click a term for the deeper article.

AFR (Annual Failure Rate): Statistical failure rate per drive per year. Consumer drives hover around ~1.5%, enterprise at 0.5-0.8%. Multiplies with drive count: 8 drives in RAID 0 = ~11.4% chance/year.
Bitrot: Silent data corruption – individual bits change unnoticed via magnetic drift, cosmic rays, or controller bugs. Classic RAID doesn't detect it. Protection: ZFS or Btrfs with block checksums. More: Bitrot & ZFS scrubbing.
CMR (Conventional Magnetic Recording): Classic HDD recording with parallel tracks, no overlap. Writes are fast and predictable. RAID-friendly. Counterpart to SMR. More: SMR vs CMR.
Hot spare: A spare drive connected to a running NAS that's automatically integrated into the RAID upon a drive failure. Counterpart to cold spare. More: Hot spare vs cold spare.
IOPS (Input/Output Operations per Second): Measure of read/write operations per second. Critical for databases and VMs. HDDs: 100-200 IOPS, SSDs: 10,000-100,000+ IOPS.
JBOD (Just a Bunch Of Disks): Multiple drives concatenated into one logical volume without RAID logic. Total capacity = sum of all drives. No protection, but on failure you only lose data on the failed drive (unlike RAID 0).
MTBF (Mean Time Between Failures): Statistical lifespan in hours. Enterprise drives spec 2-2.5 million hours MTBF, consumer 1-1.5 million. Limited usefulness – AFR is more practical.
NAS (Network Attached Storage): Storage server for home or office network. Provides storage over SMB, NFS, AFP. Examples: Synology, QNAP, TrueNAS, Unraid, DIY.
Parity: Redundancy info calculated from data on other drives. On a failure, missing data is reconstructed from it. RAID 5 uses one parity, RAID 6 two.
Rebuild / Resilver: Rebuilding a RAID after a drive swap. Data is reconstructed from parity or mirror onto the new drive. Takes 12-50 hours depending on size. ZFS calls it resilver. More: RAID rebuild time.
SHR (Synology Hybrid RAID): Synology-proprietary RAID that optimally uses mixed drive sizes. Unlike classic RAID 5/6 (limited by the smallest drive), SHR distributes data across multiple layers. SHR-1 = 1 drive protection, SHR-2 = 2.
SMART (Self-Monitoring, Analysis and Reporting Technology): Self-monitoring built into every drive. Reports values like Power_On_Hours, Reallocated_Sectors, Pending_Sectors. Read tool: smartctl. Mandatory check on used drives.
SMR (Shingled Magnetic Recording): Recording method with overlapping tracks for higher density per platter. Sustained writes are slow. Problematic in RAID and can cause rebuild failures. More: SMR vs CMR.
SSD (Solid State Drive): Solid-state storage with no moving parts. Faster, quieter, pricier per TB than HDD. In NAS often used as cache or for active data.
Striping: Splitting data into blocks distributed in parallel across drives. Boosts read/write speed. Used by RAID 0 (no protection) and RAID 5/6/10 (with protection).
TB vs TiB: TB (terabyte) = 10^12 bytes (decimal). TiB (tebibyte) = 2^40 bytes (binary, ~1.1 TB). Vendors quote TB, OS often shows TiB. A 16 TB drive = ~14.5 TiB usable.
Unraid: Commercial NAS OS by Lime Technology. Per-drive filesystem, one or two parity drives. Allows arbitrary drive sizes and easy expansion. Single-purchase license from ~$60.
UPS (Uninterruptible Power Supply): Battery buffer that cleanly shuts down a NAS on power loss. Prevents RAID corruption from unclean shutdowns. More: Best UPS for NAS.
URE (Unrecoverable Read Error): An unrecoverable read error. Statistically one URE per 10^14 bits read on consumer drives. Non-trivial during large RAID rebuilds, can abort a RAID 5. More: Rebuild time.
ZFS / RAIDZ: Modern filesystem + volume manager. Block checksums protect against bitrot. RAIDZ1/2/3 are ZFS equivalents of RAID 5/6/... with triple parity. High RAM appetite (rule of thumb: 1 GB per TB).
HexOS: Consumer-friendly NAS operating system, built on TrueNAS Scale with ZFS underneath. Goal: plug-and-play UI for home users who want ZFS benefits (RAIDZ, checksums, snapshots) without the configuration overhead. Free for home use. Under the hood it uses the same ZFS RAIDZ modes as TrueNAS.
Btrfs: Modern Linux filesystem with snapshots, checksums and built-in RAID 0/1/10. RAID 5/6 is still considered unstable – not for production. More: Btrfs RAID 5/6 issues.
Cache (SSD): SSD in front of a slower HDD pool to speed up frequent access. Read cache: speedup only; write cache: speed + buffering. ZFS distinguishes ARC (RAM) and L2ARC (SSD).
Checksum: Hash over a data block for integrity verification. ZFS and Btrfs store per-block checksums, classic RAID doesn't. Detects bitrot. More: ZFS vs ext4 vs Btrfs.
Cold spare: A reserve drive sitting on the shelf, not connected. On failure, swap manually. Saves power and wear vs hot spare. More: Hot spare vs cold spare.
Copy-on-Write (CoW): Write strategy of ZFS and Btrfs: data is never overwritten, it's written to a new location and the pointer is updated. Enables atomic snapshots and prevents the write hole.
Deduplication: Detects identical data blocks and stores them only once. ZFS dedup is very RAM-hungry (~5 GB per TB of data). Rarely worth it — snapshots/compression usually win.
ext4: Default Linux filesystem. Stable, fast, no checksums or snapshots. Works with mdadm RAID and LVM. Classic stack on OMV. More: ZFS vs ext4 vs Btrfs.
Hot swap: Swap a drive while the system is running, without shutting down the NAS. Requires a hot-swap-capable backplane and RAID mode. Standard on Synology/QNAP, hardware-dependent in DIY builds.
iSCSI: Block-level protocol: the NAS exposes a virtual disk (LUN) over the network. The client accesses it at block level and formats it itself. Typical for VMware, Hyper-V, databases.
mdadm: Linux software-RAID tool. Builds RAID 0/1/5/6/10 without a hardware controller. Solid, well-documented, but no checksums. Foundation of Synology SHR and OMV RAID.
NFS: Network File System. Unix standard for file sharing. Faster and lighter on CPU than SMB, but clunkier on Windows. A must for Linux/macOS NAS mounts.
NVMe: Modern SSD protocol over PCIe. Up to ~7000 MB/s sequential, much higher IOPS than SATA SSDs. In NAS typically used as cache or its own all-flash pool.
Pool (ZFS): ZFS storage container made of one or more vdevs. Datasets live inside the pool and share capacity dynamically. Expansion only via additional vdevs, not single drives.
RAID controller: Hardware card with its own CPU and cache that computes RAID. Pro: BBU protects against the write hole. Con: lock-in — if the controller dies you often need the same model to recover.
Scrub / Scrubbing: Background job that reads all data and compares it against checksums. Finds bitrot before it corrupts. Schedule monthly on ZFS/Btrfs. More: Bitrot & ZFS scrubbing.
SMB / CIFS: Windows file-sharing network protocol. Default on every NAS. SMB3 supports encryption and multichannel. Disable SMB1 — insecure.
Snapshot: Point-in-time copy of a dataset. On ZFS/Btrfs almost free and atomic thanks to CoW. Protects against ransomware and accidental deletion — not a backup replacement.
TBW (Terabytes Written): SSD endurance metric: total guaranteed write volume. Consumer drives ~150-600 TBW per TB, enterprise drives several PBW. Important for NAS cache SSDs.
vdev: ZFS Virtual Device: a group of drives in RAIDZ, mirror or stripe. A pool consists of one or more vdevs. IOPS scale with the number of vdevs, not with drive count.
Write hole: RAID 5/6 risk: a power loss mid-write can leave data and parity inconsistent. Mitigations: BBU on hardware controllers, ZFS/Btrfs CoW, or a UPS.
ECC RAM: Error-Correcting RAM detects and fixes single-bit flips. Strongly recommended for ZFS, since uncorrected RAM errors can be written into checksums. A must on TrueNAS servers with critical data.
HBA (Host Bus Adapter): Controller card without RAID logic — passes drives straight to the OS (IT mode). A must for ZFS and TrueNAS. Classic choice: LSI/Broadcom 9300-8i flashed to IT mode.
HDD (Hard Disk Drive): Classic spinning-platter hard drive. Cheap per TB, slow (100-250 MB/s, 100-200 IOPS), wear-prone. Default for NAS bulk storage.
10 GbE: 10 Gigabit/s networking. Only worthwhile once the pool can sustain more than 1 GbE (SSDs or 4+ HDDs striped). Switch and NIC start around ~150 EUR. More: 10 GbE NAS upgrade.
LACP / Link Aggregation: Bundles several network ports into one logical link. Increases total throughput across multiple clients (each connection still caps at a single port). Requires a managed switch.
Jumbo frames: Ethernet packets with MTU 9000 instead of 1500. Lowers CPU load on bulk transfers, mainly useful at 10 GbE+. Must be set identically on NAS, client and switch or you'll get packet loss.
VLAN: Virtual LAN. Logically separates network segments on the same switch (e.g. NAS, IoT, guests). Improves security and clarity. Requires a managed switch.
VPN: Encrypted tunnel for remote NAS access without exposing ports publicly. Recommended: WireGuard or Tailscale. Never expose SMB directly to the internet.
LVM (Logical Volume Manager): Linux layer between disks and filesystem. Lets you span volumes across multiple drives, resize live, and take snapshots. Commonly used on OMV alongside mdadm.
LUN (Logical Unit Number): Virtual disk exposed by the NAS to iSCSI or Fibre Channel clients. The client formats it like a local disk (NTFS, ext4, VMFS).
Quota: Storage cap per user, share, or dataset. Prevents one account from filling the pool. ZFS per-dataset quotas are particularly flexible.
3-2-1 Backup Rule: 3 copies of the data, on 2 different media, with 1 off-site. Industry standard — RAID doesn't count as backup, no versioning and no geographic separation.
RPO / RTO: RPO = how much data loss is acceptable (e.g. max 1 h). RTO = how fast the service must be back online after an outage. Drive your backup frequency and restore plan.
Air-gap backup: Backup medium is physically disconnected (USB drive on the shelf, rotated LTO tapes). Protects against ransomware because the backup is unreachable while the attacker is active.
LUKS / Encryption at rest: LUKS = Linux standard for disk encryption. ZFS and Btrfs ship native encryption. Protects data if a drive is stolen — not while the system is running. More: ZFS encryption guide.
Docker: Container runtime. Lets you run apps like Plex, Jellyfin, Home Assistant or Nextcloud in isolation on the NAS. Standard on Synology, QNAP, Unraid and TrueNAS Scale. More: Self-hosting on NAS.
Proxmox VE: Open-source hypervisor based on Debian. Popular combo: Proxmox on bare metal, TrueNAS or OMV as a VM. Manages KVM VMs and LXC containers from a single web UI.
Reallocated sectors: SMART attribute: number of bad sectors swapped out for reserve sectors. > 0 is an early warning, > 50 means schedule a replacement. Always check on used drives.
Workload rating: Vendor spec: maximum annual data throughput (TB/year) within warranty. Consumer HDD: 55 TB/yr, NAS HDD: 180-300 TB/yr, enterprise: 550 TB/yr. More: Disk failure statistics 2026.
Drive size mismatch: Classic RAID 5/6 uses only the smallest drive's size — larger drives waste capacity. SHR and Unraid solve this. More: Mix different drive sizes in RAID.

← Back to calculator