Hubbry Logo
Disk storageDisk storageMain
Open search
Disk storage
Community hub
Disk storage
logo
8 pages, 0 posts
0 subscribers
Be the first to start a discussion here.
Be the first to start a discussion here.
Disk storage
Disk storage
Disk storage
View on Wikipedia
from Wikipedia

Disk storage (also sometimes called drive storage) is a data storage mechanism based on a rotating disk. The recording employs various electronic, magnetic, optical, or mechanical changes to the disk's surface layer. A disk drive is a device implementing such a storage mechanism. Notable types are hard disk drives (HDD), containing one or more non-removable rigid platters; the floppy disk drive (FDD) and its removable floppy disk; and various optical disc drives (ODD) and associated optical disc media.

(The spelling disk and disc are used interchangeably except where trademarks preclude one usage, e.g., the Compact Disc logo. The choice of a particular form is frequently historical, as in IBM's usage of the disk form beginning in 1956 with the "IBM 350 disk storage unit".)

Six hard disk drives
Three floppy disk drives
A CD-ROM (optical) disc drive

Background

[edit]

Audio information was originally recorded by analog methods (see Sound recording and reproduction). Similarly, the first video disc used an analog recording method. In the music industry, analog recording has been mostly replaced by digital optical technology, where the data is recorded in a digital format with optical information.

The first commercial digital disk storage device was the IBM 350 which shipped in 1956 as a part of the IBM 305 RAMAC computing system. The random-access, low-density storage of disks was developed to complement the already used sequential-access, high-density storage provided by tape drives using magnetic tape. Vigorous innovation in disk storage technology, coupled with less vigorous innovation in tape storage, has reduced the difference in acquisition cost per terabyte between disk storage and tape storage; however, the total cost of ownership of data on disk including power and management remains larger than that of tape.[1]

Disk storage is now used in both computer storage and consumer electronic storage, e.g., audio CDs and video discs (VCD, DVD and Blu-ray).

Data on modern disks is stored in fixed length blocks, usually called sectors and varying in length from a few hundred to many thousands of bytes. Gross disk drive capacity is simply the number of disk surfaces times the number of blocks/surface times the number of bytes/block. In certain legacy IBM CKD drives the data was stored on magnetic disks with variable length blocks, called records; record length could vary on and between disks. Capacity decreased as record length decreased due to the necessary gaps between blocks.

Access methods

[edit]

Digital disk drives are block storage devices. Each disk is divided into logical blocks (collection of sectors). Blocks are addressed using their logical block addresses (LBA). Read from or write to disk happens at the granularity of blocks.

Originally the disk capacity was quite low and has been improved in one of several ways. Improvements in mechanical design and manufacture allowed smaller and more precise heads, meaning that more tracks could be stored on each of the disks. Advancements in data compression methods permitted more information to be stored in each of the individual sectors.

The drive stores data onto cylinders, heads, and sectors. The sector unit is the smallest size of data to be stored in a hard disk drive, and each file will have many sector units assigned to it. The smallest entity in a CD is called a frame, which consists of 33 bytes and contains six complete 16-bit stereo samples (two bytes × two channels × six samples = 24 bytes). The other nine bytes consist of eight CIRC error-correction bytes and one subcode byte used for control and display.

The information is sent from the computer processor to the BIOS into a chip controlling the data transfer. This is then sent out to the hard drive via a multi-wire connector. Once the data is received onto the circuit board of the drive, they are translated and compressed into a format that the individual drive can use to store onto the disk itself. The data is then passed to a chip on the circuit board that controls the access to the drive. The drive is divided into sectors of data stored onto one of the sides of one of the internal disks. An HDD with two disks internally will typically store data on all four surfaces.

The hardware on the drive tells the actuator arm where it is to go for the relevant track, and the compressed information is then sent down to the head, which changes the physical properties, optically or magnetically, for example, of each byte on the drive, thus storing the information. A file is not stored in a linear manner; rather, it is held in the best way for quickest retrieval.

Rotation speed and track layout

[edit]
Comparison of several forms of disk storage showing tracks (not-to-scale); green denotes start and red denotes end.
* Some CD-R(W) and DVD-R(W)/DVD+R(W) recorders operate in ZCLV, CAA or CAV modes.

Mechanically, there are two different motions occurring inside the drive. One is the rotation of the disks inside the device. The other is the side-to-side motion of the head across the disk as it moves between tracks.

There are two types of disk rotation methods:

  • constant linear velocity (used mainly in optical storage) varies the rotational speed of the optical disc depending upon the position of the head, and
  • constant angular velocity (used in HDDs, standard FDDs, a few optical disc systems, and vinyl audio records) spins the media at one constant speed regardless of where the head is positioned.

Track positioning also follows two different methods across disk storage devices. Storage devices focused on holding computer data, e.g., HDDs, FDDs, and Iomega zip drives, use concentric tracks to store data. During a sequential read or write operation, after the drive accesses all the sectors in a track, it repositions the head(s) to the next track. This will cause a momentary delay in the flow of data between the device and the computer. In contrast, optical audio and video discs use a single spiral track that starts at the innermost point on the disc and flows continuously to the outer edge. When reading or writing data, there is no need to stop the flow of data to switch tracks. This is similar to vinyl records, except vinyl records started at the outer edge and spiraled in toward the center.

Interfaces

[edit]

The disk drive interface is the mechanism/protocol of communication between the rest of the system and the disk drive itself. Storage devices intended for desktop and mobile computers typically use ATA (PATA) and SATA interfaces. Enterprise systems and high-end storage devices will typically use SCSI, SAS, and FC interfaces in addition to some use of SATA.

Basic terminology

[edit]
Disk
Generally refers to magnetic media and devices.
Disc
Required by trademarks for certain optical media and devices.
Platter
An individual recording disk. A hard disk drive contains a set of platters. Developments in optical technology have led to multiple recording layers on DVDs.
Spindle
The spinning axle on which the platters are mounted.
Rotation
Platters rotate; two techniques are common:
  • Constant angular velocity (CAV) keeps the disk spinning at a fixed rate, measured in revolutions per minute (RPM). This means the heads cover more distance per unit of time on the outer tracks than on the inner tracks. This method is typical with computer hard drives.
  • Constant linear velocity (CLV) keeps the distance covered by the heads per unit time fixed. Thus the disk has to slow down as the arm moves to the outer tracks. This method is typical for CD drives.
Track
The circle of recorded data on a single recording surface of a platter.
Sector
A segment of a track
Low level formatting
Establishing the tracks and sectors.
Head
The device that reads and writes the information—magnetic or optical—on the disk surface.
Arm
The mechanical assembly that supports the head as it moves in and out.
Seek time
Time needed to move the head to a new position (specific track).
Rotational latency
Average time, once the arm is on the right track, before a head is over a desired sector.
Data transfer rate
The rate at which user data bits are transferred from or to the medium. Technically, this would more accurately be entitled the "gross" data transfer rate.

See also

[edit]

References

[edit]
Revisions and contributorsEdit on WikipediaRead on Wikipedia

Disk storage

View on Grokipedia
from Grokipedia
Disk storage is a class of secondary storage technology that uses one or more rotating disks to store and retrieve digital data, enabling random access to information through electronic, magnetic, or optical recording on the disk surface. The primary mechanism involves platters—rigid or flexible disks coated with magnetic material—where data is represented as magnetized spots or patterns, read and written by heads positioned on mechanical arms. This design provides non-volatile persistence, meaning data remains intact without power, distinguishing it from primary memory like RAM, and supports capacities from megabytes in early models to terabytes in modern systems. The foundational development of disk storage began in the mid-20th century, with IBM's 1956 release of the RAMAC 305, the first commercial (HDD), featuring 50 platters storing 5 million characters in a refrigerator-sized unit. Key components include the spindle for rotation (typically 5,400–15,000 RPM), tracks as concentric data rings, and sectors as the smallest addressable units (usually 512 bytes or 4 KB). Access involves seek time for head movement, rotational latency for disk positioning, and transfer time for data movement, with overall performance optimized by scheduling algorithms like Shortest Seek Time First (SSTF) to minimize delays in multi-request environments. Historically, disk storage encompassed magnetic variants such as floppy disks—introduced in the for portable, low-capacity (up to 1.44 MB) storage using flexible Mylar platters—and optical disks like CDs (1983, 650–700 MB) and DVDs (1995, up to 17 GB), which use laser-based reading for read-only or rewritable media. In contemporary , HDDs remain dominant for high-capacity, cost-effective bulk storage (e.g., up to 36 TB per drive as of 2025), while solid-state drives (SSDs) using have emerged as faster alternatives without moving parts, though they are not traditional disk storage due to the absence of rotation. Disk storage's evolution has driven advancements in file systems, configurations for redundancy, and data centers, underpinning everything from personal to .

History and Development

Early Innovations

The invention of magnetic drum memory in 1932 by Austrian engineer Gustav Tauschek marked a significant precursor to modern disk storage, utilizing a rotating coated with magnetic to store and retrieve via fixed read-write heads. Tauschek's prototype, patented while working for an subsidiary in , demonstrated the feasibility of random-access magnetic storage on a cylindrical surface, influencing later flat-disk designs by adapting principles of persistent, high-speed retention. This technology addressed limitations of earlier electrostatic and mercury-delay-line memories, providing capacities up to several kilobytes with access times in milliseconds, though drums remained bulky and power-intensive. The breakthrough to disk-based storage occurred in 1956 with the development of the IBM 350, the world's first commercial random-access disk drive, as part of the system. Led by IBM engineer , the team adapted magnetic recording techniques originally derived from audio technologies—such as iron oxide coatings on rotating surfaces—to create a stack of 50 aluminum platters, each 24 inches in diameter, capable of storing 5 megabytes of data in total. Johnson's innovations, including precise head positioning over the platters, enabled average access times of about 600 milliseconds, revolutionizing data handling for business applications like by allowing random retrieval without sequential tape scanning. Early disk storage faced substantial challenges, including exorbitant costs, mechanical fragility, and compatibility with vacuum-tube . The IBM 350 RAMAC system was rented for $3,200 per month—equivalent to over $35,000 today—making it accessible only to large enterprises despite its modest capacity. Mechanical issues were pronounced, as high-speed rotation at 1,200 rpm caused platters to warp, necessitating solutions like gluing pairs of disks together for stability, while read-write heads floated on a thin air cushion just 800 microinches above the surface to prevent crashes. Integration with vacuum-tube-based computers like the 305 added complexity, as the drive's electronics had to interface with heat-generating, unreliable tubes prone to frequent failures. To mitigate seek-time delays inherent in moving-head designs like the , fixed-head disks emerged in the , employing one read-write head per track on large single-platter units up to 3 feet in diameter. This configuration eliminated radial head movement, reducing access latencies to tens of milliseconds and improving reliability in early environments, though at the expense of higher costs and limited scalability per unit. These innovations laid the groundwork for subsequent advancements, including the shift toward removable disk packs in the following decade.

Transition to Digital Storage

The transition to digital storage in disk technology accelerated in the early 1960s with the introduction of removable disk packs, enabling greater portability and flexibility compared to fixed-platter systems. The IBM 1311 Disk Storage Drive, announced on October 11, 1962, was the first commercial random-access disk drive featuring interchangeable disk packs, each containing six 14-inch platters with a capacity of approximately 2 million characters (roughly 2 megabytes, depending on encoding mode). This innovation addressed limitations in earlier fixed-disk designs by allowing users to swap packs for data transport and security, marking a shift toward more practical, user-managed storage solutions in mainframe environments. By the , the rise of s spurred demand for compact, affordable storage, leading to smaller form factors that democratized disk access beyond large-scale mainframes. A pivotal development was the 8-inch , invented by an team led by and introduced in 1971 as a read-only medium for loading into mainframe controllers. This flexible magnetic disk, with an initial capacity equivalent to about 80 kilobytes, facilitated easier data exchange in minicomputer systems and paved the way for subsequent read-write versions commercialized by starting in 1973. Concurrently, areal density in rigid disk drives advanced dramatically, from 2,000 bits per square inch in the 1956 IBM 350 RAMAC to over 1 million bits per square inch by the mid-1980s, driven by innovations like thin-film inductive heads introduced in 's 3370 drive in 1979. These heads, fabricated using for precise thin metallic layers, enabled closer head-to-disk spacing and higher recording densities without excessive wear. A landmark advancement came in 1973 with IBM's 3340 "" disk drive, which adopted a sealed head-disk assembly to minimize airborne contamination—a persistent issue in open-pack designs. This architecture integrated low-mass, low-load landing heads with lubricated platters in an enclosed module, supporting capacities up to 35 megabytes per spindle while improving reliability and enabling further gains. By protecting the recording surfaces from dust and particles, the Winchester design reduced error rates and maintenance needs, facilitating the production of smaller, more robust drives suitable for diverse computing applications. These technological shifts had profound market implications, eroding the dominance of storage by the 1980s as disks offered vastly superior times—typically milliseconds versus minutes for tape rewinding. While tape remained cost-effective for archival backups, its sequential nature proved inadequate for the interactive workloads of emerging personal and , prompting a migration to disk-based systems for primary storage. This transition not only boosted overall storage capacities but also accelerated the integration of disks into broader digital ecosystems.

Core Principles

Basic Terminology

In disk storage systems, particularly hard disk drives (HDDs), a platter refers to a rigid, circular disk typically made of aluminum or and coated with a thin magnetic film on both surfaces to enable through patterns. Each platter is mounted on a central spindle and spins at a constant rotational speed, allowing multiple platters to be stacked in a single drive for increased capacity. The head, or read-write head, is an electromagnetic positioned at the end of an actuator arm that hovers microns above the platter surface to read or write data by detecting or altering magnetic fields without physical contact. One head is dedicated to each recording surface of , and all heads move in unison across the stack to access data. Data on a platter is organized into tracks, which are concentric circular paths etched on the magnetic surface at a fixed radius from the center, where bits are stored sequentially along the ring. Each track is further subdivided into sectors, the smallest addressable units of storage, traditionally holding 512 bytes of data, though modern drives may use 4 KB sectors for improved efficiency. Sectors include headers, data fields, and error-correcting codes to ensure reliable access. A consists of the set of tracks at the same radial position across all platters in the drive, forming a vertical alignment that allows simultaneous access by all heads without radial movement. This organization minimizes seek operations when data spans multiple surfaces. Seek time measures the duration required for the actuator arm to position the heads from their current location to the target track or on a platter. It encompasses track-to-track seek time, typically around 1-2 ms for adjacent tracks, and average seek time, which ranges from 5-10 ms depending on the drive's and distance traveled.

Disk Geometry and Data Layout

Disk geometry refers to the physical arrangement of data storage areas on a disk platter, typically organized into concentric tracks subdivided into sectors. In hard disk drives (HDDs), platters rotate at a (CAV), maintaining a fixed rotational speed regardless of the radial position, which results in higher linear velocities at outer tracks compared to inner ones. This CAV approach simplifies mechanical design and control but leads to varying data transfer rates across the disk surface. In contrast, constant linear velocity (CLV) varies the rotational speed to keep linear speed constant, a method more typical in devices like CDs and DVDs rather than HDDs. To optimize storage capacity under CAV, modern HDDs employ zoned bit recording (ZBR), which groups tracks into radial zones where each zone maintains a constant number of sectors per track but adjusts the sector count to achieve approximately constant linear bit density. Outer zones contain more sectors than inner zones due to their larger circumference, maximizing areal density without exceeding magnetic recording limits on inner tracks. ZBR was first commercially implemented in 1961 by Bryant Computer Products in their 4000 Series disk drives, and it became a standard feature in high-capacity HDDs by the to support increasing storage demands. The initial physical layout is created through low-level formatting, a factory process that magnetically encodes tracks, sectors, servo patterns for head positioning, and headers/trailers on the platter surface, typically defining sectors of 512 bytes or 4 KB. High-level formatting follows, performed by the operating system or user, to overlay logical structures such as the , partition tables, and metadata (e.g., FAT for legacy systems or for Windows), without altering the underlying physical sectors. Access to data is abstracted from physical geometry via (LBA), which treats the disk as a linear array of consecutively numbered blocks starting from 0, hiding complexities like zoning and variable sector counts. LBA was introduced as a standard in the ATA-1 specification in 1994, using a 28-bit to support up to 137 GB, with later extensions like 48-bit LBA enabling petabyte-scale capacities. Defects in the media, such as manufacturing flaws or wear-induced errors, are managed through techniques like sector slipping and spare sectors to ensure reliability. Sector slipping "skips" defective sectors during formatting by remapping subsequent logical blocks to shift data forward, effectively bypassing the bad area without gaps in addressing; this is often combined with vertical error-correcting codes to detect and handle errors efficiently. Additionally, disks reserve spare sectors (typically 1-5% of total capacity) within zones or cylinders to replace defective ones transparently, with the drive updating the defect list and remapping accesses on-the-fly.

Mechanical and Access Mechanisms

Read-Write Operations

In hard disk drives, read-write heads employ a design where the slider floats above the rotating platter surface at a precise clearance, typically 3 to 10 nanometers, maintained by an generated from the induced by the platter's . This nanoscale proximity enables high-density access while minimizing wear, and during spin-down or power-off, the heads are parked using ramp loading mechanisms that lift the slider onto an inclined ramp at the platter's outer edge to prevent contact with the recording surface. The write process involves passing an through a coil in the inductive write head, which generates a localized strong enough to align the magnetic domains—small regions of aligned atomic moments—on the platter's ferromagnetic coating in a desired orientation representing . To optimize storage efficiency and mitigate inter-symbol interference, data is encoded using Run-Length Limited (RLL) schemes, such as (2,7)-RLL, which constrain the minimum and maximum run lengths of consecutive zeros between transitions, allowing up to 67% more bits per inch compared to earlier methods like . During the read process, the head senses changes in magnetic flux from the passing domains; early designs used inductive sensors that detect voltage induced by flux variations, but since the 1990s, magnetoresistive technologies have dominated for greater sensitivity. (GMR) heads, introduced commercially in 1997, exploit the quantum mechanical effect where electrical resistance in multilayered ferromagnetic/non-magnetic structures varies significantly with applied magnetic fields, enabling detection of weaker signals from denser recordings and supporting areal densities over 10 Gb/in². Later advancements include (TMR) heads, introduced in 2004, which use a tunnel barrier for even greater resistance changes, enabling higher areal densities over 1 Tb/in² in contemporary drives as of 2025. To ensure , error correction employs Reed-Solomon codes embedded in servo sectors and user data, capable of correcting multiple symbol errors and achieving post-correction bit error rates below 10^{-12} in typical magnetic recording channels. Overwriting data poses challenges due to residual magnetism from incomplete domain realignment, potentially allowing partial recovery of prior bits, but this is mitigated through techniques like applying (AC) erasure fields that randomize magnetic orientations without net alignment, effectively reducing to negligible levels.

Rotation and Track Management

The spindle motor in hard disk drives maintains a constant rotational speed, typically ranging from 5,400 RPM in some consumer models to 10,000 RPM in certain high-performance enterprise drives, with most modern drives at 7,200 RPM as of 2025, to ensure consistent data access timing. This results in rotational latency, the time required for the desired sector to rotate under the read-write head, which averages half of one full ; for a 7,200 RPM drive, this equates to approximately 4.16 milliseconds. Track following relies on servo mechanisms embedded within the disk , where servo wedges—radial sectors containing position error signals—are strategically placed to provide periodic feedback for precise head alignment. These embedded servo patterns enable closed-loop feedback control, allowing the voice coil motor to make fine adjustments and maintain the head on the target track with sub-micron accuracy during . During access operations across different zones of the disk, adaptive flying height management, often implemented via thermal fly-height control (TFC), dynamically adjusts the head's protrusion to optimize clearance; this is particularly crucial in inner zones where linear velocities are lower, helping to prevent head crashes by maintaining a stable nanometer-scale gap between the head and platter surface. Transfer rates vary significantly between inner and outer zones due to differences in linear at constant RPM, with outer zones achieving higher speeds; for example, the Seagate Exos X24 (as of 2023) sustains up to 285 MB/s at the outer diameter, decreasing to lower rates (around 150–200 MB/s) at the inner diameter. To mitigate rotational and access latencies, modern hard disk drives incorporate onboard DRAM caching, typically 256 MB to 512 MB in capacity, which buffers data from sequential reads ahead of time, enabling faster retrieval from cache if subsequent requests align with the prefetched blocks.

Interfaces and Integration

Historical Standards

The development of disk storage interfaces began with proprietary systems tailored to early mainframe computers. In 1956, introduced the 305 RAMAC system, which featured the Model 350 disk storage unit as its core component. This interface was a custom, cable-based connection designed specifically for integration with mainframes such as the 305 and later models like the 650 and 1401. It operated at a low data transfer rate of 8.8 KB/s, reflecting the era's technological constraints and focus on rather than high-speed throughput. As personal computing emerged in the late 1970s, interfaces shifted toward more standardized and accessible designs for smaller systems. introduced the ST-506 interface in 1980 alongside its namesake 5 MB , marking a pivotal step for early PCs. This parallel interface used two 34-pin ribbon cables—one for control signals and one for data—employing (MFM) encoding to achieve a transfer rate of 5 Mbit/s (approximately 0.625 MB/s). The follow-up ST-412 model in 1981 doubled capacity to 10 MB and was adopted by for the PC/XT, solidifying the interface's role in establishing 5.25-inch form factors as an industry norm. The 1980s saw the rise of more versatile standards to support multiple devices and higher performance. The Enhanced Small Device Interface (ESDI), developed in the early and formalized as ANSI X3.170 in 1990, acted as a bridge between simpler interfaces like ST-506 and more advanced protocols. It utilized separate 20-pin data and 34-pin control cables, supporting transfer rates starting at 10 Mbit/s and scaling up to 24 Mbit/s (about 3 MB/s) in later implementations, which enabled its use in minicomputers and high-end workstations from vendors like and . ESDI improved on prior designs by incorporating embedded servo data for better track following, though it still required dedicated controllers. A landmark standardization effort culminated in the Small Computer System Interface (SCSI), approved as ANSI X3.131 in 1986. This parallel bus architecture allowed daisy-chaining of up to 7 devices (8 total including the host) on a single cable, with SCSI-1 specifying an 8-bit bus at 5 MB/s transfer speed using asynchronous or synchronous modes. Subsequent variants evolved the standard: SCSI-2 (1990) added command queuing and wider buses for up to 15 devices, while SCSI-3 (late 1990s) introduced serial options and speeds exceeding 320 MB/s in parallel forms. SCSI's command set enabled broad compatibility across peripherals, influencing server and ecosystems. Despite their innovations, these historical interfaces faced notable limitations that constrained scalability and reliability. , for instance, required unique device IDs (0-7 for narrow variants) to arbitrate bus access, leading to conflicts and if duplicates occurred, often necessitating careful configuration. Cabling posed another challenge: bulky, shielded parallel ribbons (e.g., 50-pin Centronics-style) limited cable lengths to 6 meters and introduced issues in daisy-chained setups, while address limitations capped total devices without expanders. These factors contributed to the transition toward serial interfaces in later decades.

Contemporary Protocols

Serial ATA (SATA), introduced in 2003 as a successor to parallel ATA, represents a pivotal shift to serial interfaces for consumer and prosumer disk storage, enabling higher data transfer rates and improved efficiency. The SATA 3.0 specification, finalized in 2009, supports speeds up to 6 Gb/s, facilitating faster access to large storage volumes in personal computers and workstations. Key features include hot-swapping, which allows devices to be connected or disconnected without system shutdown, and Native Command Queuing (NCQ), which optimizes command execution by handling up to 32 simultaneous operations to reduce overhead and improve throughput. For enterprise environments, (SAS) emerged in 2004 as a robust serial protocol tailored for high-reliability storage systems, offering dual-port and beyond consumer needs. As of 2025, the SAS-4 standard (INCITS 519-2014, revised 2018), achieves transfer rates of 22.5 Gb/s, with SAS-5 (INCITS 554-2023) introducing further enhancements for hyperscale applications. SAS employs expanders to connect up to 65,536 devices in a single domain theoretically, enabling expansive storage arrays while maintaining with drives for cost-effective hybrid deployments. This compatibility allows SAS hosts to seamlessly integrate peripherals, broadening its applicability in mixed environments without requiring separate cabling infrastructures. Fibre Channel (FC) serves as the backbone for storage area networks (SANs), providing high-bandwidth, low-latency connectivity for enterprise disk storage over extended distances. As of 2025, the Gen 7 (64 Gb/s) protocol, defined in FC-PI-7, delivers speeds up to 64 Gb/s using optical or electrical links, with optical transceivers supporting reaches of up to 10 km on single-mode fiber; Gen 8 (128 Gb/s) standards are finalized, with products expected by late 2025. This capability is essential for distributed data centers, where FC enables block-level access to disk arrays across fabrics, ensuring consistent performance in mission-critical applications like and database clustering. FC's and fabric services further enhance and manageability in large-scale SAN topologies. NVMe over Fabrics (NVMe-oF) extends the NVMe protocol—originally designed for low-latency SSDs—across networked fabrics, with applicability to high-end HDDs in hybrid storage setups to leverage efficient queuing mechanisms. By emulating PCIe-style command submission and completion queues over transports like Ethernet, , or , NVMe-oF minimizes latency compared to traditional protocols, achieving sub-millisecond response times for remote disk access. This supports scalable, disaggregated storage pools, allowing HDDs in enterprise arrays to benefit from NVMe's parallelism without the physical constraints of direct-attached PCIe lanes. As of 2025, NVMe-oF has seen widespread adoption in cloud and AI workloads, with enhancements for (RoCE) improving efficiency in large-scale deployments. Contemporary protocols also incorporate advanced power management to address energy efficiency in always-on storage systems. SATA's DevSleep feature, introduced in the 3.1 specification, enables devices to enter an ultra-low-power idle state by powering down the PHY layer and associated circuitry, consuming as little as 5 mW while maintaining rapid wake-up times under 10 ms. This complements partial and slumber modes, reducing overall power draw in laptops and data centers by optimizing idle periods without compromising accessibility. Similar efficiencies are integrated into SAS and FC standards, promoting sustainable operation in power-sensitive deployments.

Types and Applications

Magnetic Hard Disks

Magnetic hard disk drives (HDDs) consist of one or more rigid platters coated with magnetic material, stacked on a central spindle that rotates at high speeds, typically 5,400 to 15,000 RPM, to enable access. The platters are housed in a sealed to minimize and maintain stable conditions, with read-write heads mounted on arms that position them precisely over tracks. The assembly is driven by a motor, which uses electromagnetic forces to rapidly move the arms across the platters, allowing seek times as low as 3-5 milliseconds in modern designs. Capacity in magnetic HDDs has evolved dramatically, from the IBM 3380 model in 1980, which exceeded 1 GB per drive as the first to break that barrier, to over 20 TB in enterprise units by 2023, reaching 32 TB in enterprise units as of 2025. This growth stems from advances in areal density, achieved through technologies like perpendicular magnetic recording (PMR) in the and more recent innovations such as (HAMR) by Seagate, which enables 20-24 TB drives shipping in 2023, with 32 TB drives beginning to ship in 2025, and energy-assisted perpendicular magnetic recording (ePMR) combined with (SMR) by , supporting capacities up to 32 TB as of 2024 without thermal lasers. These methods allow bits to be written more densely by temporarily altering the magnetic of the media, pushing beyond the superparamagnetic limit of traditional recording. Common form factors for magnetic HDDs include the 3.5-inch size, predominant in desktop computers and for its balance of capacity and cooling, and the 2.5-inch variant for laptops, offering portability with thicknesses of 7-9.5 mm. Enterprise environments favor 2.5-inch drives in 15 mm heights for dense server racks, enabling higher storage per unit volume in data centers while maintaining compatibility with standard bays. Reliability in magnetic HDDs is quantified by (MTBF), typically rated at 1-2.5 million hours for enterprise models, reflecting projected operational lifespan under continuous use. (S.M.A.R.T.) enhances this by continuously tracking attributes like error rates, temperature, and spin-up time, issuing predictive alerts when thresholds indicate impending failure, though it cannot foresee all issues. In applications, magnetic HDDs serve as bulk storage in data centers, where (SMR) boosts density by overlapping tracks like , achieving up to 20-25% higher capacity than conventional methods but incurring sequential write penalties due to the need to rewrite entire bands for updates. This makes SMR ideal for write-once, read-many workloads like archiving, reducing through lower cost per terabyte.

Removable Disk Formats

Removable disk formats encompass portable magnetic and optical media designed for easy interchangeability between devices, evolving from early flexible disks to higher-capacity cartridges in the late . These formats prioritized user accessibility for transfer and , using flexible or rigid magnetic coatings within protective enclosures to store on spinning platters or discs. Unlike fixed hard drives, removable disks allowed physical transport of , though they typically offered lower capacities and slower access speeds due to their emphasis on portability. The earliest prominent removable format was the 8-inch floppy disk, developed by in 1971 to load for the System/370 and its 3330 disk storage controller, with the first units shipped that year for the System/370. This single-sided disk, using flexible magnetic media coated with iron oxide, provided an initial formatted capacity of 80 KB, equivalent to about 3,000 punched cards, making it a revolutionary alternative to tape or card-based input for mainframe data loading. By the mid-1970s, it supported double-sided operation for up to 256 KB in some variants, but its large size limited it to professional and industrial applications. Following the 8-inch model, the 5.25-inch floppy disk emerged in 1976 from as the "Minifloppy" drive, targeting minicomputers and early personal systems. Initial single-density versions offered around 110 KB unformatted, but double-density (DD) formats standardized at 360 KB formatted capacity became common by the early 1980s for PC compatibility. High-density (HD) evolution in the mid-1980s pushed this to 1.2 MB, using enhanced magnetic coatings and error correction to support operating systems like and . These disks, still flexible but housed in softer sleeves, facilitated widespread data sharing in the nascent personal computing era. The 3.5-inch , developed by in 1980 and standardized in 1982 by the Microfloppy Industry Committee, marked the pinnacle of flexible magnetic for consumer use. Enclosed in a rigid shell for durability, it initially offered 400 KB in double-density but achieved 1.44 MB in high-density formats by 1984, compatible with PCs and Macintosh systems. This format's smaller size and shutter mechanism improved reliability, dominating data exchange until the 1990s with billions produced for and file backups. Beyond floppies, cartridge-based magnetic formats like the Bernoulli drive, introduced by in 1982, used air-bearing technology to suspend the read-write head above a flexible 8-inch or 5.25-inch disk, preventing crashes and enabling capacities from 20 MB to 150 MB by the late 1980s. This design, inspired by principles, targeted professional backup needs with removable cartridges up to 230 MB in later iterations. Similarly, 's , launched in 1994, provided a more affordable cartridge alternative with initial 100 MB capacity on 3.5-inch-like media, scaling to 250 MB and 750 MB versions by the early 2000s; it became a staple for storage, outselling floppies briefly. The Jaz drive followed in 1996, offering 1 GB per cartridge in a rigid, shock-resistant , later expanding to 2 GB, though reliability issues like the "click of death" from head crashes tempered its adoption. Optical variants of removable disk storage, such as the developed jointly by and , debuted in 1982 with a read-only capacity of 650 MB on a 120 mm disc using laser-based reading. Standardized for data in 1983, it enabled mass distribution of software and archives, far exceeding magnetic floppies in density due to pit-based encoding rather than magnetic domains. While primarily read-only, writable formats like emerged later, but 's interchangeability relied on the standard from 1988, ensuring cross-platform compatibility on PCs and workstations. The decline of removable disk formats accelerated post-2000 with the rise of USB flash drives, which offered solid-state capacities starting at 128 MB—surpassing Zip and Jaz—without mechanical parts, at lower costs and higher speeds. Floppy production ceased entirely by 2010 when , the last major manufacturer, halted output due to negligible demand, though cartridges like Zip persisted in niche markets until the mid-2000s. Today, these formats endure primarily for archival purposes in legacy industrial systems, such as embroidery machines and aviation controls, where compatibility trumps modern alternatives.

References

  1. https://www.sciencedirect.com/topics/computer-science/disk-technology
  2. https://pages.cs.wisc.edu/~remzi/OSTEP/file-disks.pdf
  3. https://homepage.cs.uri.edu/faculty/wolfe/book/Readings/Reading05.htm
  4. https://www.computerhistory.org/timeline/memory-storage/
  5. https://www.cs.uic.edu/~jbell/CourseNotes/OperatingSystems/10_MassStorage.html
  6. https://www.ibm.com/think/topics/hard-disk-drive-vs-solid-state-drive
  7. https://www.seagate.com/news/seagate-ships-30tb-drives-to-meet-global-surge-in-data-center-ai-storage-demand/
  8. https://www.computerhistory.org/storageengine/tauschek-patents-magnetic-drum-storage/
  9. https://www.computerhistory.org/revolution/memory-storage/8/252
  10. https://www.ibm.com/history/ramac
  11. https://www.ibm.com/history/rey-johnson
  12. https://www.computerhistory.org/storageengine/first-commercial-hard-disk-drive-shipped/
  13. https://www.slashgear.com/1971341/ibm-first-hdd-hard-disk-drive-1956/
  14. https://www.engr.colostate.edu/~sudeep/wp-content/uploads/CD6.14-P374493.pdf
  15. https://digitalcommons.njit.edu/fac_pubs/20488/
  16. https://www.computerhistory.org/storageengine/disk-packs-offer-portability-security/
  17. http://archive.computerhistory.org/resources/text/Oral_History/IBM_1311_2311/IBM_1311_2311.oral_history.2005.102657931.pdf
  18. https://www.computerhistory.org/storageengine/floppy-disk-loads-mainframe-computer-data/
  19. https://www.ibm.com/history/floppy-disk
  20. https://www.computerhistory.org/storageengine/thin-film-heads-introduced-for-large-disks/
  21. https://www.computerhistory.org/storageengine/winchester-pioneers-key-hdd-technology/
  22. https://d1yx3ys82bpsa0.cloudfront.net/groups/ds-ibm-3340.pdf
  23. https://dl.acm.org/doi/10.1145/3708997
  24. https://www.ecs.csun.edu/~cov/comp222s16/notes/notes_harddrive.pdf
  25. https://opendsa-server.cs.vt.edu/ODSA/Books/CS3/html/Diskdrive.html
  26. https://people.eecs.berkeley.edu/~kubitron/courses/cs162-F15/fa15/static/lectures/17.pdf
  27. https://www.gartner.com/en/information-technology/glossary/clv-constant-linear-velocity
  28. https://www.cs.cmu.edu/~harchol/Papers/JSA2011.pdf
  29. https://www.computerhistory.org/storageengine/hdd-competitors-enter-the-market/
  30. https://www.techtarget.com/whatis/definition/logical-block-addressing-LBA
  31. https://www.cs.cmu.edu/~natassa/courses/15-721/papers/IEEEComputer.DiskModel.pdf
  32. https://cml.berkeley.edu/wp-content/uploads/gold-reports/G07002.pdf
  33. https://documents.westerndigital.com/content/dam/doc-library/en_us/assets/public/western-digital/collateral/white-paper/loadunload-white-paper-final.pdf
  34. https://web.eece.maine.edu/~vweaver/classes/ece571_2022f/ece571_lec21.pdf
  35. https://yogeshmodhe.files.wordpress.com/2016/01/chm_topic-2_notes_storage-devices-and-its-interfacing2.pdf
  36. https://d1yx3ys82bpsa0.cloudfront.net/groups/magnetoresistive-heads.pdf
  37. https://nationalmaglab.org/magnet-academy/read-science-stories/science-simplified/giant-magnetoresistance/
  38. https://cnrs.hal.science/hal-03882390/document
  39. https://www.cs.auckland.ac.nz/~pgut001/pubs/secure_del.html
  40. https://www.seagate.com/content/dam/seagate/en/content-fragments/products/datasheets/exos-x24/exos-x24-DS2080-2307US-en_US.pdf
  41. https://courses.cs.washington.edu/courses/cse378/10sp/lectures/lec26.pdf
  42. https://ieeexplore.ieee.org/document/6971641
  43. https://ieeexplore.ieee.org/document/6971515
  44. https://ieeexplore.ieee.org/document/5297575/
  45. https://www.seagate.com/content/dam/seagate/assets/support/internal-hard-drive/enterprise-hard-drives/exos-x24/_shared/files/Seagate_EXOS24_CMR_ISE_SED%2810-12-16-20-24TB%29_Rev-C.pdf
  46. https://www.seagate.com/content/dam/seagate/migrated-assets/www-content/product-content/barracuda-fam/barracuda-new/files/multi-tier-caching-technology-white-paper-2017.pdf
  47. https://community.cadence.com/cadence_blogs_8/b/breakfast-bytes/posts/ramac-park
  48. https://gunkies.org/wiki/ST506_disk_interface
  49. https://www.ardent-tool.com/storage/ESDI_Evolution.html
  50. http://bitsavers.org/pdf/micropolis/ESDI_spec_X3.170-90.pdf
  51. https://blog.ansi.org/ansi/incits-131-1994-2013-computer-systems-interface/
  52. https://nvlpubs.nist.gov/nistpubs/Legacy/FIPS/fipspub131.pdf
  53. https://www.cablewholesale.com/support/technical_articles/scsi_unscrewed.php
  54. https://www.sweetwater.com/sweetcare/articles/scsi-frequently-asked-questions-faq/
  55. https://www.seagate.com/files/docs/pdf/evolution-to-sata-6gb-storage-tp610.1-1001us.pdf
  56. https://sata-io.org/sites/default/files/documents/SATA-Revision-3.0-Press-Release-FINAL-052609.pdf
  57. https://sata-io.org/system/files/specifications/SerialATA_Revision_3_1_Gold.pdf
  58. https://www.seagate.com/support/kb/serial-attached-scsi-frequently-asked-questions-210751en/
  59. https://www.t10.org/drafts.htm
  60. https://blog.ansi.org/ansi/incits-554-2023-sas-protocol-layer-5-spl-5/
  61. https://h10032.www1.hp.com/ctg/Manual/c00256909.pdf
  62. https://fibrechannel.org/wp-content/uploads/2024/12/6Introducing-128G-Fibre-Channel-for-Storage-Networking.pdf
  63. https://www.cisco.com/c/en/us/products/collateral/storage-networking/mds-9396t-32-gbps-96-port-fibre-channel-switch/data-sheet-c78-740633.html
  64. https://nvmexpress.org/wp-content/uploads/NVMe_Over_Fabrics.pdf
  65. https://www.purestorage.com/au/knowledge/what-is-nvme-over-fabrics-nvme-of.html
  66. https://sata-io.org/sites/default/files/documents/SATA-IO-DEVSLP-Press-Release_FINAL.pdf
  67. https://sata-io.org/sites/default/files/documents/SATADevSleep-and-RTD3-WP-037-20120102-2_final.pdf
  68. https://investor.wdc.com/static-files/65bbf704-73ac-4e81-9139-2b774229d66a
  69. https://aphelis.net/first-gigabyte-hard-drive-ibm-3380/
  70. https://www.computerhistory.org/storageengine/heat-assisted-magnetic-recording-hamr-finally-arrives/
  71. https://www.westerndigital.com/company/newsroom/press-releases/2024/2024-10-15-western-digital-delivers-worlds-highest-capacity-epmr-hdds
  72. https://investor.wdc.com/static-files/f89af5fe-52a8-47b4-a1d5-2631d5e3c4a6
  73. https://www.seagate.com/www-content/product-content/enterprise-hdd-fam/enterprise-capacity-2-5-hdd/en-us/docs/ent-capacity-2-5-hdd-ds1719-6-1505gb.pdf
  74. https://www.seagate.com/www-content/datasheets/pdfs/exos-x-12-DS1946-1-1709US-en_US.pdf
  75. https://www.seagate.com/support/kb/my-system-reported-a-smart-error-on-the-drive-184619en/
  76. https://documents.westerndigital.com/content/dam/doc-library/en_us/assets/public/western-digital/collateral/white-paper/white-paper-shingled-magnetic-recording-hdd-technology.pdf
  77. https://www.computerhistory.org/storageengine/minicomputers-stimulate-floppy-disk-sales/
  78. https://computer.howstuffworks.com/floppy-disk-drive1.htm
  79. https://obsoletesony.substack.com/p/how-sony-invented-the-35-inch-floppy
  80. https://www.backblaze.com/blog/history-removable-computer-storage/
  81. https://www.silicon.co.uk/data-storage/storage/tales-tech-history-iomega-zip-drive-209499
  82. https://www.cnet.com/tech/mobile/iomega-beats-up-on-syquest/
  83. https://www.computerhistory.org/storageengine/philips-demonstrates-digital-compact-disc/
  84. https://www.wired.com/story/why-the-floppy-disk-just-wont-die/
  85. https://brainstation.io/magazine/floppy-disks-its-been-a-great-30-years
Add your contribution
Related Hubs
User Avatar
No comments yet.