What is data storage?

Storage options and setups range from a single device to large-scale, distributed systems. For everyday users, data storage houses apps, photos and media on phones and personal devices. In enterprise settings, it supports everything from document archiving to backup and disaster recovery. Some data is stored briefly and discarded, while other data is kept for weeks, years or indefinitely depending on its purpose.

Data volume and complexity continue to grow, driven in part by big data analytics. According to a report from the International Data Corporation (IDC), the amount of data created, captured and replicated worldwide is on pace to nearly triple from 2025 to 2029. Much of this growth comes from artificial intelligence (AI), machine learning (ML) and Internet of Things (IoT) workloads. These technologies need storage optimization to scale without driving up cost.

Data storage security protects data on-premises and in the cloud against data breaches, cyberattacks and other threats. As ransomware attacks continue to rise, storage is no longer just where data lives—it has become the enforcement layer for operational resilience and compliance.

Modern storage security features include encryption, access permissions, data masking and automated audit reporting. According to the IBM Cost of a Data Breach Report 2025, the global average cost of a data breach was USD 4.4 million. Cyber resilience, an organization’s ability to prevent, withstand and recover from cybersecurity incidents, has become a core part of meeting that challenge. It combines business continuity and disaster recovery (BCDR), information security and organizational resilience to help organizations withstand and recover from attacks.

Healthcare, finance and government face especially strict requirements for data recovery and retention. Regulations, such as the EU’s Digital Operational Resilience Act (DORA) for financial institutions and HIPAA’s data security requirements for healthcare data, require organizations to regularly test recovery processes and document the results. Such testing mandates make clear that storage itself must be resilient, not only the systems around it.

For industries that must preserve records and maintain data integrity, immutable storage offers an added layer of protection. For example, immutable snapshots—unchangeable, point-in-time copies of data—prevent stored data from being changed or deleted for a set or indefinite period, guarding against tampering, cyberattacks and ransomware.

In simple terms, modern computers or terminals connect to storage devices either directly or through a network. Users instruct computers to manage data access from, and store data on, these storage devices. Today, much of that data is machine generated, created automatically from devices like sensors, applications and other systems.

Data storage is different from computer memory, such as random access memory (RAM). While RAM gives the central processing unit (CPU) fast access to data it’s actively working with, that data disappears once the computer powers down. Storage retains it. Users can save data to a device and retrieve it whenever they need it. Storage can also be shared across multiple users and devices.

Data storage comes in two broad categories: primary storage and secondary storage. Primary storage holds the data and instructions the CPU needs most often, and CPUs can access primary storage directly. It sits close to the CPU, which makes it fast to read and write.

Secondary (persistent) storage holds data the CPU accesses less frequently. CPUs cannot access this data directly—data must first be copied into primary storage. This storage offers far more capacity for long-term retention.

For a deeper comparison, see “Primary storage vs. secondary storage: what’s the difference?”

Persistent storage can be organized in three main ways:

- File storage
- Block storage
- Object storage

File storage is the type that most everyday computer users will be most familiar with—when a user saves something to their computer’s hard disk drive (HDD) or solid-state drive (SSD), they are using file storage. Files can also be saved to network-attached storage (NAS) devices, in cloud-based file storage or on portable media such as flash drives.

In file storage, data is saved as complete files, with an extension such as .jpg or .txt that indicates its format. Files are organized in a hierarchical structure of folders and subfolders, and users access a file by entering its path or navigating to it using the folder hierarchy.

Common use cases for file storage include personal files, shared team storage, media and content production, and software development environments.

Block storage, sometimes called block-level storage, divides data into fixed-size chunks called blocks. These blocks are stored separately and identified by a unique address. Because blocks do not have an inherent file system and carry only minimal metadata, they can be formatted and managed to fit the needs of a given application. This makes block storage adaptable to a wide range of workloads.

Block storage is often used for databases and virtual machines because it enables fast, efficient access to data. When a database needs to update a record, it can address and rewrite the specific blocks that contain the targeted data (known as “in-place updates”), rather than rewrite an entire file. This “direct pipeline” to data provides applications with precise, low-latency control over where and how data is written.

Object storage, often called object-based storage, organizes data as self-contained units called objects. Objects are stored in a flat structure, typically organized into containers called buckets, rather than the hierarchical folders of file storage or the addressed blocks of block storage. Each object includes the data itself, metadata and a unique identifier, and this identifier is what enables the storage system to retrieve objects without a hierarchical structure. It is highly scalable and often used for large volumes of unstructured data.

Users access and manage objects through application programming interfaces (APIs) rather than file paths or direct disk access. Object storage generally has higher latency than block or file storage, which makes it best suited for data that must be kept but is rarely accessed, such as archives of legal, medical or financial documents. And because it stores massive amounts of data economically, with metadata that makes objects easy to search, it’s often used to hold the datasets for training AI models.

File storage, block storage and object storage are all widely used today, often within the same enterprise environment. Organizations with mixed workloads might use file storage for shared documents, block storage for databases and object storage for archives. For a closer look at how organizations manage all three, see “What is unified storage?”

Storage architectures fall into two main categories: direct-attached storage (DAS) and network-based storage.

Direct-attached storage (DAS) connects to a single computer without a network in between, which limits its usefulness for data sharing. DAS devices include HDDs and SSDs—both internal and external—and USB flash drives.

Older media also falls into this category, including floppy disks, optical discs (CDs, DVDs and Blu-ray discs) and magnetic tape, often grouped into reels, or tape libraries.

Network-based storage lets multiple computers access shared storage over a network, making it well suited for collaboration. Because storage is independent of any single machine—and can even be located off-site—it also supports backups and protects against data loss.

Two standard network-based storage architectures are network-attached storage (NAS) and storage area networks (SANs). NAS is often a single device containing multiple drives, typically configured as a redundant array of independent disks (RAID) for data protection.

A SAN—sometimes called the network behind the servers—is a specialized, high-speed network that connects servers to storage devices and presents that storage at the block level so it appears to servers like locally attached disks. SANs often use Fibre Channel, a data transfer technology built for fast, high-volume, highly reliable connections.

Organizations typically turn to SANs as server demands grow and workloads require fast, reliable access to shared storage.

Here’s how NAS and SAN differ:

NAS

- Single storage device
- File storage
- TCP/IP Ethernet network
- Suited to fewer simultaneous users
- Slower performance
- Limited scalability
- Lower cost, simpler setup

SAN

- Network of multiple devices
- Block storage
- Typically Fibre Channel (or iSCSI) network
- Handles many simultaneous users
- Faster performance
- Highly scalable
- Higher cost, more complex setup

Storage technologies span physical media like SSDs and HDDs, deployment models like cloud and hybrid cloud, and systems purpose-built for workloads such as AI and edge computing.

- SSD and flash storage
- Hybrid storage
- Cloud storage
- Hybrid cloud storage
- AI storage
- Edge storage

Data storage has increasingly moved toward software-defined approaches that make data management more agile and efficient. A report from Technavio estimates the global software-defined storage (SDS) market will grow by more than USD 105 billion from 2024 to 2028. 

Here’s a breakdown of these technologies.

- Software-defined storage (SDS)
- Storage virtualization
- Hyperconverged storage
- Intelligent storage

SDS is a type of data storage in which a software layer decouples storage resources from the underlying hardware. It uses virtualization to create a unified pool of storage that can be managed manually through a dashboard or allocated automatically via API. 

Unlike traditional NAS or SAN systems, SDS can scale and adapt without new hardware. It also streamlines storage management by automating tasks such as provisioning, monitoring and troubleshooting.

Storage virtualization pools physical storage resources from multiple systems so that they appear as a single, logically managed pool of storage. SDS goes a step further and abstracts the storage services from the hardware itself. Administrators manage SDS through a central console, simplifying oversight of security, reliability and efficiency across the entire storage environment.

Hyperconverged storage integrates storage directly into a hyperconverged infrastructure (HCI) stack alongside computing and networking. Because HCI virtualizes these resources rather than tying them to individual pieces of hardware, hyperconverged storage is more flexible and scalable than traditional storage.

Intelligent storage uses automation and machine learning to improve storage efficiency, manageability and performance. Typical capabilities include predictive analysis, autonomous data tiering, automatic software updates and proactive troubleshooting—all within a unified management system.