Data storage

Data storage is the persistent or temporary representation of digital information on a physical or logical medium. Unlike random-access memory, which normally loses its contents when power is removed, persistent storage is designed to retain data across shutdowns, although every medium has a finite lifetime and failure rate.⁴ Storage is commonly organized in a hierarchy: processor caches and memory provide low-latency access; local solid-state or magnetic drives provide working capacity; and networked or removable media support sharing, backup, and archiving. At the software level, block storage presents addressable units to an operating system, while file storage organizes named files and directories. Object storage instead keeps data with metadata and identifiers, a model suited to large, distributed repositories.

Capacity is measured in bytes, while performance may be described by latency, throughput, and input/output operations per second. These measures are not interchangeable: a high-capacity tape library can be economical for archives but unsuitable for interactive applications.

Storage media differ in how they encode bits and in the kinds of access they support. Hard disk drives record magnetic patterns on rotating platters and remain important for comparatively inexpensive, high-capacity storage. Solid-state drives store charge states in flash memory, eliminating moving parts and generally reducing access latency, but flash cells wear as they are programmed and erased; controllers therefore use wear leveling, spare capacity, and error correction.¹

Optical discs use laser-readable marks and can be useful for distribution or offline retention, while magnetic tape offers high capacity and low energy use when stored without continuous access. Cloud services combine disks, flash, and sometimes tape behind network interfaces, hiding the physical medium from users. The Google File System and later cloud architectures demonstrated how software can combine ordinary machines with replication and monitoring to provide a larger logical store.⁵

Reliable storage depends on redundancy, integrity checking, monitoring, and recovery procedures rather than on any single indestructible device. A RAID array can mirror or stripe data across drives and may reconstruct information after a component failure, but RAID is not a backup: accidental deletion, malware, fire, or a shared software error can affect every copy. Distributed systems may use replication or erasure coding, which divides data into fragments and adds calculated fragments so missing pieces can be rebuilt with less overhead than full replication.⁵

Administrators also distinguish availability from durability. A service can be reachable yet return corrupted or incomplete data, so systems employ checksums, versioning, snapshots, immutable copies, and regular restoration tests. NIST guidance treats storage infrastructure as a security boundary requiring asset inventories, access management, isolation, logging, patching, and tested continuity plans.¹ Encryption protects confidentiality, but key loss can make perfectly intact data unusable.

Long-term storage is a migration problem as much as a media problem. Digital preservation must retain not only bits but also usable formats, metadata, provenance, and the software or technical context needed to interpret them; the Library of Congress therefore distinguishes preservation planning from simply keeping files on a server.³ A file can survive on a readable disk yet become inaccessible when its format, encoding, or dependencies disappear.

Offline and air-gapped copies have a niche role because they limit the reach of ransomware and network failures, although they introduce handling and recovery costs. Tape libraries may use automated robotic loading, and archival systems often apply write once, read many controls to prevent alteration. Government records programs emphasize scheduled transfer, integrity verification, and retention decisions rather than indefinite accumulation. Storage media can also fail silently through bit rot, so fixity checks and geographically separate copies are central to trustworthy preservation.

Block storage

Storage that exposes fixed-size addressable blocks to a host, which can then create its own file system or database layout.

File storage

Storage that organizes data as named files in directories and presents operations such as create, read, write, and delete.

Object storage

A storage model in which data, metadata, and a unique identifier are managed as an object, commonly through a network interface.

RAID

Redundant Array of Independent Disks; techniques that combine multiple drives for redundancy, performance, or both.

Erasure coding

A method that adds calculated fragments to data so lost fragments can be reconstructed without keeping a complete duplicate of every fragment.

Fixity

Evidence that a digital object has not changed, commonly established by comparing a later checksum with an earlier recorded value.