The world's first hard disk, the IBM RAMAC 305, released in 1956, could hold only 5 MB of data, weighed 970 kg, and its dimensions were comparable to an industrial refrigerator. Modern corporate flagship models boast capacities of up to 20 TB. Just imagine: 64 years ago, to store that amount of information, it would have taken over 4 million RAMAC 305 units, and the data center required to house them would have exceeded 9 square kilometers, while today a small box weighing around 700 grams will suffice! Much of this incredible increase in storage density has been achieved through advancements in magnetic recording methods.
It’s hard to believe, but the fundamental design of hard drives has remained unchanged for nearly 40 years, starting from 1983: that was when the first 3.5-inch hard drive, the RO351, developed by the Scottish company Rodime, was introduced. This little device came with two magnetic platters of 10 MB each, meaning it could store twice as much data as the updated 5.25-inch ST-412 released by Seagate in the same year for IBM 5160 personal computers.
Rodime RO351 — the first 3.5-inch hard drive in the world
Despite its innovative design and compact size, upon its release, the RO351 turned out to be practically unwanted, and all subsequent attempts by Rodime to establish itself in the hard drive market failed, leading the company to cease operations in 1991, selling virtually all its assets and drastically reducing its workforce. However, it was not destined for bankruptcy: soon after, major hard drive manufacturers began approaching it to acquire licenses to use the patented Scottish form factor. Today, 3.5 inches is an established standard for both consumer HDDs and enterprise-class storage.
With the advent of neural networks, Deep Learning, and the Internet of Things (IoT), the volume of data generated by humanity has been growing exponentially. According to estimates by the analytical agency IDC, by 2025, the amount of information generated by both individuals and the devices around us is expected to reach 175 zettabytes (1 zettabyte = 10^21 bytes). This is in stark contrast to 2019, when it was 45 zettabytes, to 2016, which saw 16 zettabytes, and back in 2006, the total volume of data produced throughout history was a mere 0.16 zettabytes. Modern technologies, including advanced data storage methods, are helping to manage this information explosion.
LMR, PMR, CMR, and TDMR: what is the difference?
The principle of operation of hard drives is quite simple. Thin metallic platters coated with a layer of ferromagnetic material (a crystalline substance capable of retaining magnetization even in the absence of an external magnetic field at temperatures below the Curie point) move at high speed relative to the recording heads (5400 revolutions per minute or more). When an electric current is applied to the write head, an alternating magnetic field is created, changing the direction of the magnetization vector of the domains (discrete areas of the substance) of the ferromagnet. Data reading occurs either through the phenomenon of electromagnetic induction (the movement of domains relative to the sensor induces an alternating electric current in the sensor) or via the giant magnetoresistive effect (the sensor's electrical resistance changes under the influence of a magnetic field), as implemented in modern drives. Each domain encodes one bit of information, taking a logical value of "0" or "1" depending on the direction of the magnetization vector.
For a long time, hard drives used Longitudinal Magnetic Recording (LMR), where the magnetization vector of the domains lay in the plane of the magnetic platter. Despite its relative simplicity of implementation, this technology had a significant drawback: to overcome coercivity (the transition of magnetic particles to a single-domain state), a substantial buffer zone had to be left between tracks (the so-called guard space). As a result, the maximum recording density achieved at the end of this technology was only 150 Gbit/in².
By 2010, LMR had been nearly completely replaced by Perpendicular Magnetic Recording (PMR). The main difference between this technology and longitudinal magnetic recording is that the magnetic orientation vector of each domain is positioned at a 90° angle to the surface of the magnetic platter, significantly reducing the spacing between tracks.
As a result, the data recording density has significantly increased (up to 1 Tbit/in² in modern devices) without sacrificing speed characteristics or the reliability of hard drives. Currently, perpendicular magnetic recording is dominant in the market, often referred to as CMR (Conventional Magnetic Recording). It is important to understand that there is no difference between PMR and CMR; they are simply two different terms for the same technology.
When studying the technical specifications of modern hard drives, you may also encounter the enigmatic abbreviation TDMR. This technology is mainly used in enterprise-class storage devices. . From a physics perspective, TDMR (short for Two Dimensional Magnetic Recording) is no different from the familiar PMR: as before, we deal with non-overlapping tracks, with domains oriented perpendicular to the magnetic plates. The difference between the technologies lies in the approach to reading information.
In the block of magnetic heads of hard drives made using TDMR technology, each write head is paired with two read sensors that read data from each track simultaneously. This redundancy allows the HDD controller to effectively filter out electromagnetic noise, which is caused by intertrack interference (ITI).
Solving the ITI problem provides two extremely important advantages:
- the reduction of interference allows for increased recording density by decreasing the distance between tracks, resulting in an overall capacity gain of up to 10% compared to conventional PMR;
- in combination with RVS technology and a three-position micro-actuator, TDMR effectively counters rotational vibration caused by hard drive operation, helping maintain stable performance levels even in the most challenging operating conditions.
What is SMR and how is it used?
The size of the write head is approximately 1.7 times larger than that of the read sensor. This significant difference is easily explained: if the writing module were made even smaller, the magnetic field strength it could generate would be insufficient to magnetize the domains of the ferromagnetic layer, meaning that data simply would not be retained. This issue does not arise with the read sensor. Furthermore, its miniaturization helps further reduce the impact of the aforementioned ITI on the information reading process.
This fact forms the basis of Shingled Magnetic Recording (SMR). Let’s understand how it works. When using traditional PMR, the write head is offset relative to each previous track by a distance equal to its width plus the width of the guard space.
When using the shingled magnetic recording method, the write head shifts forward by only part of its width, meaning that each previous track is partially overwritten by the subsequent one: the magnetic tracks overlap like roof shingles. This approach additionally increases recording density, providing up to a 10% capacity gain, without affecting the reading process. For example, one can mention — the world's first 3.5-inch drives with a capacity of 20 TB featuring SATA/SAS interfaces, made possible by a new magnetic recording technology. Thus, switching to SMR disks allows for denser data storage in the same spaces while minimizing IT infrastructure upgrade costs.
Despite this significant advantage, SMR also has an obvious drawback. Since the magnetic tracks overlap, updating data requires rewriting not only the specific fragment but all subsequent tracks within the magnetic platter, which can exceed 2 terabytes, potentially leading to severe performance drops.
This issue is addressed by grouping a certain number of tracks into distinct clusters called zones. Although this approach to data storage organization slightly reduces the overall HDD capacity (as sufficient gaps must be maintained between zones to prevent overwriting tracks from neighboring groups), it significantly speeds up the data updating process, since only a limited number of tracks are now involved.
Shingled magnetic recording allows for several implementation options:
- Drive Managed SMR (SMR, managed by the drive)
Its primary advantage is the lack of need to modify the host's software and/or hardware, as the HDD controller takes over the data writing process. Such drives can be connected to any system that has the necessary interface (SATA or SAS), after which the storage will be ready for use.
The drawback of this approach lies in the variability of performance levels, making Drive Managed SMR unsuitable for enterprise applications where consistent system performance is critical. However, such drives perform well in scenarios that allow sufficient time for background data defragmentation. For example, DMSMR drives , optimized for use in small NAS systems with 8 bays, will be an excellent choice for an archiving or backup system that requires long-term storage of backups.
- Host Managed SMR (HM-SMR)
Host Managed SMR is the preferred implementation of zoned recording for use in an enterprise environment. In this case, the host system manages data flows and read/write operations using extensions of the ATA interface (Zoned Device ATA Command Set, ZAC) and SCSI (Zoned Block Commands, ZBC), developed by the INCITS T10 and T13 committees.
When using HM-SMR, the entire available capacity of the drive is divided into two types of zones: Conventional Zones, which are used for storing metadata and random writes (essentially acting as a cache), and Sequential Write Required Zones, which occupy the majority of the hard drive's total capacity where data is written strictly sequentially. Unordered data is saved in the caching area and can later be moved to the corresponding sequential write zone. This allows all physical sectors to be written sequentially in a radial direction and rewritten only after a cyclic transfer, achieving stable and predictable system performance. Furthermore, HM-SMR disks support random read commands similarly to drives using standard PMR.
Host Managed SMR is implemented in enterprise-class hard drives .
The lineup includes high-capacity SATA and SAS drives designed for use in hyperscale data centers. Support for Host Managed SMR significantly expands the applicability of such hard drives: in addition to backup systems, they are well-suited for cloud storage, CDNs, or streaming platforms. The high capacity of the hard drives allows for a substantial increase in storage density (in the same racks) at minimal upgrade costs, while low power consumption (no more than 0.29 watts per terabyte of stored data) and heat generation (on average 5 °C lower than that of competitors) additionally reduce operational costs for data center maintenance.
The only drawback of HMSMR is its relatively complex implementation. The fact is that, to date, no operating system or application can work with such drives "out of the box", which requires significant changes in the software stack to adapt the IT infrastructure. This primarily concerns the operating system itself, which is quite a non-trivial task in modern data centers that use multi-core and multi-socket servers. To learn more about options for implementing Host Managed SMR support, visit the specialized resource , dedicated to the issues of zoned data storage. The information collected here will help you preliminarily assess the readiness of your IT infrastructure for transitioning to zoned storage systems.
- Host Aware SMR
Devices with Host Aware SMR combine the convenience and flexibility of Drive Managed SMR with the high write speed of Host Managed SMR. Such drives are backward compatible with legacy storage systems and can operate without direct control from the host; however, in this case, as with DMSMR drives, their performance becomes unpredictable.
Similar to Host Managed SMR, Host Aware SMR employs two types of zones: Conventional Zones for random writes and Sequential Write Preferred Zones. The latter, unlike the aforementioned Sequential Write Required Zones, automatically convert to a regular status if unordered data writing begins.
Host Aware SMR implementation includes internal mechanisms for recovery from unordered writes. Unordered data is written to cache areas from which the disk can transfer information to the sequential write zone after all necessary blocks have been received. To manage unordered writes and perform background defragmentation, the disk uses an indirect access table. However, if enterprise applications require predictable and optimized performance, this can only be achieved when the host takes full control of all data streams and write zones.
Source: habr.com
