RAID

RAID
Name	RAID
Caption	Redundant Array of Independent Disks
Invented	1980s
Inventor	David A. Patterson; Garth A. Gibson; Randy Katz
Type	Data storage virtualization

RAID is a data storage virtualization technology that combines multiple physical disk drives into one or more logical units for redundancy, performance, or both. Originating from academic research at the University of California, Berkeley, RAID has been adopted across enterprise environments including vendors such as IBM, Dell Technologies, Hewlett-Packard Enterprise, and NetApp. RAID concepts have influenced standards and products from organizations like the Storage Networking Industry Association and have been discussed in venues such as the ACM and IEEE conferences.

Overview

RAID groups multiple physical drives into arrays managed by hardware controllers or software stacks like those in Linux, Microsoft Windows, FreeBSD and macOS. Early formalization appeared in a 1988 paper by researchers at University of California, Berkeley that contrasted performance and reliability trade-offs, influencing designs from manufacturers including Seagate Technology, Western Digital, and Hitachi Global Storage Technologies. Implementations range from embedded controllers in servers from Sun Microsystems to virtualization platforms such as VMware ESXi and Microsoft Hyper-V.

Levels and Variants

Common RAID configurations include nested and standard levels labeled with integers and letters originating from the Berkeley report and later industry practice. Examples are RAID 0, RAID 1, RAID 5, RAID 6, and nested forms like RAID 10 (mirrored stripes) used by vendors like EMC Corporation and NetApp. Proprietary and hybrid variants—often created by companies such as IBM, Dell EMC, and HP—include implementations marketed as RAID-Z in products from Sun Microsystems/Oracle Corporation (ZFS lineage) and erasure-coding schemes used by hyperscalers such as Google and Amazon Web Services. Academic extensions and alternative taxonomies have been proposed at conferences like USENIX and published in journals associated with the ACM and IEEE.

Implementation and Architecture

RAID can be implemented in hardware controllers produced by companies like Adaptec and LSI Corporation, or in software within operating systems such as Linux (mdadm), FreeBSD (geom), and Microsoft Windows (Storage Spaces). Architectures vary: controller-based solutions integrate with firmware from vendors like Broadcom and utilize interfaces such as SATA and SAS; software-based solutions interact with kernel subsystems and file systems such as ZFS and Btrfs. Cloud providers including Google Cloud Platform and Amazon Web Services use distributed storage architectures and erasure coding that conceptually relate to RAID principles but integrate with object stores like Google Cloud Storage and Amazon S3. Management ecosystems include tools and standards from SNIA and monitoring integrations with platforms like Nagios and Prometheus.

Performance and Reliability

Performance characteristics depend on level, disk type (from suppliers like Samsung Electronics and Micron Technology), and workload patterns described in benchmarks run by organizations such as SPEC. RAID 0 improves throughput but offers no redundancy; RAID 1 provides mirroring for availability often used in systems built by Dell Technologies and HP Enterprise; RAID 5 and RAID 6 use parity schemes with trade-offs examined in studies at Carnegie Mellon University and Massachusetts Institute of Technology. Reliability is influenced by phenomena studied by researchers at National Institute of Standards and Technology and appears in failure analyses from large operators such as Backblaze. Rebuild times, mean time between failures (MTBF) figures from manufacturers like Seagate Technology and Western Digital, and the impact of bit rot addressed by file systems from Oracle Corporation (ZFS) are central to resilience planning.

Use Cases and Deployment

RAID is used across enterprise storage arrays from EMC Corporation, SAN deployments integrating Fibre Channel switches from Brocade Communications Systems, and NAS appliances from NetApp and Synology Inc.. Typical deployments include database servers running Oracle Database and Microsoft SQL Server, virtualization hosts using VMware ESXi and KVM, and media servers leveraged by post-production houses using software from Adobe Systems. Cloud and edge deployments adapt RAID concepts when designing fault-tolerant local caches or persistent volumes in orchestration platforms such as Kubernetes and OpenStack.

Limitations and Alternatives

RAID is limited by rebuild performance, correlated failures documented by providers like Backblaze, and by newer storage paradigms. Alternatives and complements include erasure coding used by Hadoop Distributed File System and object stores like Amazon S3, replication strategies in distributed databases such as Apache Cassandra and MongoDB, and file systems designed for integrity such as ZFS and Btrfs. Emerging storage media technologies from companies like Intel Corporation (Optane) and research from institutions such as Stanford University and MIT influence future directions, while standards bodies including SNIA and IEEE continue to publish guidance shaping adoption.

Category:Data storage