Solid state drive
| Storage medium Solid State Drive (SSD)
|
|
 Various SSDs (Computex Ritek RiDATA Ultra-S) |
|
| General | |
|---|---|
| Type | Semiconductor component |
| capacity | up to 100 TB (Nimbus Data ExaDrive DC100, March 2018) |
| lifespan |
Write operations per cell: 1,000 ( TLC in 21 nm production) 3,000 ( MLC in 25 nm production) 5,000 (MLC in 34 nm production) 10,000 (MLC in 50 nm production) 100,000 ( SLC in 50 nm production) nm production) up to 5 million (selected SLC chips) |
| size | different |
| origin | |
| Launch | 1978 by StorageTek |
| predecessor | Hard disk drive |
A solid-state drive or a solid-state drive (short SSD ; from the English borrowed more rarely), semiconductor drive or solid-state memory called, is a non-volatile data storage of computer technology . The designation Drive ( English for drive ) refers to the original and usual definition for this medium of computers; however, unlike conventional designs, SSDs contain no moving parts. The design and the electrical connections can correspond to the standards for drives with magnetic or optical storage disks , but do not have to. You can, for example, as a PCIe - plug-in card to be executed. When a magnetic disk (. English Disk Drive Hard , HDD) with a solid-state combined -Speicher to a device is called a hybrid hard drive (Engl. Hybrid hard drive, HHD, nor English. Solid state hybrid drive, SSHD).
Solid-state drives were developed over the course of the second half of the 20th century until they became capable of mass use beyond individual applications. Their costs were initially very high in relation to the storage capacity, but just like the sizes with the same performance quickly decreased in accordance with Moore's Law , so that they also became economically viable for special uses around the turn of the millennium. Nevertheless, the prices for SSDs (in euros per gigabyte) in July 2018 were still several times the price of a conventional magnetic storage drive.
Due to the lack of moving parts, SSDs are mechanically more robust than conventional drives. They also have shorter access times and work silently. Despite their robustness, they can fail or cause system errors. The cause are mostly errors in the firmware , which is always coming onto the market in an immature form and is only improved later through updates.
term
In electronics, the term “solid state” means that semiconductor components are used. This distinguishes them from other storage technologies such as core memories , punch cards or memories with moving mechanical parts such as rotating magnetic disks. In analogy to drive technologies such as HDDs , FDDs and ODDs , the medium is referred to as "Drive".
Development and history
Early SSDs
Solid-state drives have their origin in the 1950s with two similar technologies, magnetic core storage and Charged Capacitor Read-Only Storage (CCROS), an early form of read -only storage . These supporting forms of storage appeared in the era of electronic tube computers, but were then given up again with the advent of cheaper drum storage systems .
In the 1970s and 1980s, SSDs were implemented in the semiconductor memory of the early supercomputers from IBM , Amdahl and Cray , but were rarely used because of their very high price. In the late 1970s, General Instruments brought the Electrically Alterable ROM (EAROM, another form of read-only memory) onto the market, which was very similar to the later NAND flash technology. However, since the lifespan of these memories was less than ten years, the technology was abandoned by many companies. In 1976 Dataram began selling a product called Bulk Core , which provided up to 2MB of solid state memory compatible with Digital Equipment Corporation (DEC) and Data General (DG) computers. In 1978, Texas Memory Systems introduced a 16- kilobyte RAM solid-state drive for oil production companies to use to record seismic data. In the following year (1979) StorageTek developed the first RAM solid state drive.
The Sharp PC-5000 , introduced in 1983, used 128 kilobyte solid-state cartridges that operated with magnetic bubble memory . In 1984 Tallgrass Technologies Corporation presented a 40 MB backup unit with an integrated 20 MB SSD that could alternatively be used as a drive. In September 1986, Santa Clara Systems announced the BatRam: a 4 MB mass storage system that could be expanded up to 20 MB. The system included rechargeable batteries in order to be able to supply power to the chip even if the current flow was interrupted. In 1987 the EMC Corporation built SSDs in mini computers for the first time , but stopped this development again in 1993.
Flash-based SSDs
In 1983 the Psion MC 400 Mobile Computer was delivered with four slots for removable storage in the form of flash-based solid-state disks. These slots were the same type used on the Psion Series 3 for flash memory cards. The major disadvantage of these modules was that they had to be formatted each time in order to free up space from deleted or modified files. Old versions of files that were deleted or edited continued to take up space until the module was formatted.
In 1991, SanDisk released a 20MB solid state drive that sold for $ 1,000 . In 1995, M-Systems first introduced a flash-based solid-state drive that did not require batteries to receive the data. However, it wasn't as fast as DRAM-based solutions. From that point on, SSDs have been successfully used as HDD replacements by military and aerospace organizations.
In 1999, BiTMICRO introduced a number of products in the field of flash-based SSDs, including an 18 GB 3.5-inch SSD. In 2007 Fusion-io introduced a PCIe-based SSD with a performance of 100,000 IOPS in a single card with a capacity of up to 320 GB. In 2009, a Flash SSD from OCZ Technology was presented at CeBIT , with a capacity of one terabyte (using a PCIe x8 interface), a maximum write speed of 654 MB / s and a maximum read speed of 712 MB / s. In December of the same year, Micron Technology announced an SSD that would use a 6- gigabit SATA interface.
Enterprise flash storage
Enterprise flash drives (EFDs, enterprise = English for companies) were designed for applications that require high IOPS performance and work reliably and efficiently. In most cases, an EFD is an SSD with a more extensive set of specifications than a standard SSD. The term was first used by EMC in January 2008 to identify SSD manufacturers that provided products with these higher standards. However, there are no standards or rules that distinguish EFDs and SSDs from one another, which is why in principle every manufacturer can state that they produce EFDs.
In 2012, Intel introduced the SSD DC S3700 - an EFD designed to deliver consistent performance. Little attention was paid to this field before.
Procedure
Two types of memory chips are used: Flash- based and SDRAMs .
Flash memories are particularly energy-efficient and even power-independent when it comes to retaining the content. The manufacturer promises around ten years.
SDRAM chips are volatile and consume significantly more energy per gigabyte than a conventional hard drive. Your advantage is the significantly higher speed. Also introduced as “ RAM disks ” in the mid-1990s , they were used right from the start in servers where caches , temporary files and journals from file, web, database servers, etc. were filed. They can be implemented as a plug-in card or as a device with an emulated hard disk interface - often with a backup battery or a separate power connection. Finding any data is 700 times faster than with a hard drive. They are 80 times as fast as flash technology. A second advantage is the almost unlimited rewritability, which is similar to a hard disk; Flash chips are limited to 100,000 to 5 million write cycles here. This restriction applies to individual flash cells. When worn, this can often be automatically replaced with a reserve cell using SMART .
It makes sense to combine the speed of SDRAMs with the data retention of other types of storage - permanent storage. For example, some manufacturers also integrate a conventional hard drive into the housing of the SDRAM SSD, for example to have a backup in the event of a power failure. Conversely, more and more SDRAM and flash chips have been built into conventional hard disks as temporary storage (so-called “ cache ”).
Areas of application
The elimination of the sensitive motor bearings and read / write mechanics of the drives with rotating disks results in increased shock resistance. The temperature tolerance is also greater. This applies to both the temperature range in which SSDs can be operated and the tolerance with regard to temperature changes. Both points qualify SSDs for mobile use. Flash-based SSDs are therefore most often found in MP3 players and USB sticks . Because the price per GB has fallen significantly in the meantime, they are now also interesting for (sub) notebooks . Another advantage is the constant, compared to rotating disks, very low latency in accessing the stored data. While accessing data in physically far apart sectors on a hard disk requires a relatively long time to move the read head to the new position (similar to a record player ), with an SSD you can use the address of a data block regardless of the previously relevant block the information contained is read or written. In addition, in contrast to hard disks, the sequential transfer rates do not decrease when the form factor is reduced (with hard disks the outer tracks of larger disks have a larger circumference, which means there is more data space and more data can be read per rotation). Hybrid hard drives and pure SSD models have been on the market since 2007.
In stationary use, SDRAM-based SSDs are more likely to be found , and mostly far from the mass market. They are used in applications that work with heavy drive loads ( databases , sorting applications ) by repeatedly requesting small amounts of data from a wide variety of storage areas. These drives are also often used by developers and testers to measure the performance of hard disk controllers and buses, since they use them to the maximum. Its history began in 1978 when the StorageTek company brought the "Solid State Disk STK 4305" onto the market, which was compatible with the IBM 2305 hard disk drive and was used with System / 370 mainframes . StorageTek itself uses the term "solid-state disk".
With the increasing performance of flash SSDs and better controllers, a comparably fast NAND- based drive was available for the first time in 2008 . It is offered as a PCIe- x4 card for the benefit of faster system connection, which is why it cannot be used as a drive for the operating system, as the system must first start in order to be able to address the "fusion io" via a driver . That and the price of 50 euros per gigabyte make it uninteresting for the end customer market , but represent a very cheap offer for a performance at RamSan level in the above-mentioned area of application.
In 2015, storage capacities of up to around 10 TB were offered; the design of these models is PCIe2.0 x16.
Solid-state drives are also often used in the area of embedded systems , in which the only thing that matters is not using mechanical parts. A one-chip microcontroller application often does not have a hard disk connector for space and energy reasons; instead, their control program or operating system is usually in a flash chip. Every PC also has one, it contains the firmware (such as EFI or BIOS ).
Other areas of application can be found in environments in which dirt, vibrations and pressure fluctuations, temperature and magnetic fields (space travel) prevent the use of mechanical plates.
SSDs in the end customer market
SSDs are in the process of supplementing or replacing conventional hard disk technology, initially especially in mobile, but now also in stationary devices. With the replacement by flash memory, numerous distinguishing features between manufacturers will disappear. These include the points of volume and cooling requirement, but also the very similar shock resistance and access time due to the principle involved . The manufacturers have freedom of design in terms of speed, capacity, reliability, price, energy requirement, housing size and weight, accessories and other features (e.g. encryption).
Hybrid hard drive / SSHD
Function and technology
The hybrid hard drive combines a conventional hard drive with a much smaller solid-state memory. Its small size is intended to absorb the additional price, but make its advantages accessible to a broad market.
DDR SDRAM
The combination with DDR-SDRAM is currently only offered by one manufacturer within Japan and far from the mass market from around 1000 euros. The DTS “Platinum HDD” uses a chip from the same manufacturer, which should learn over time which content is recommended for the fast buffer storage. This retains its data through a capacitor for one and a half minutes after the power supply has ended and consists of a 1 gigabyte DDR SDRAM module. It is housed together with a 2.5 ″ hard disk in a 3.5 ″ housing. As a result, this approach is not suitable for mobile devices, but saves a third of the energy of conventional 3.5 ″ hard drives. Since a chip makes the selection here, this drive speeds up any operating system; with HHDs, the operating system has to take on this task. So far only Windows Vista and Windows 7 can do this . In the desktop and small server area, the drive for data volumes under one gigabyte can significantly outperform any flash drive. The built-in hard disk holds between 80 and 200 GB. However, even with "normal" hard disks, DDR / DDR2-SDRAM is sometimes used as cache , but only a maximum of 128 MB.
Flash memory
The combination with Flash is becoming more and more widespread due to the support of major manufacturers as well as mobile suitability and data retention. There are two technical implementations. Intel does not integrate the flash memory into the hard disk itself, but uses a proprietary connection on the mainboard , just like for the main memory . This does not actually create a hybrid hard drive , but the effect achieved is the same. Intel calls this principle “ Turbo Memory ”. All other providers of this technology are hard drive manufacturers and integrate the flash memory into the drive itself - mostly 256 MB. Intel uses four to eight times the capacity.
Both variants are based on the fact that flash chips can deliver their data with less delay than the hard drive itself. The SDRAM buffers already present in the hard disks lose their content without a permanent power supply.
However, when it comes to writing, Flash is not only slower than this SDRAM, it also undercuts the hard disk itself. It is therefore not a replacement, but an addition. A file is therefore not included in the flash area when it is accessed for the first time, but only after frequent use; sometimes only individual components. When reading, these are then made available much faster than the hard disk could. This is only started when required, i.e. for files that are not used much. When working on the Internet or in the office, the hybrid concepts are therefore often silent and very energy-saving (around 0.3 W). These two points, together with the higher shock resistance at standstill, are their advantages. Since these are particularly beneficial for mobile use, HHDs have so far only been manufactured in 2.5 inches. Thanks to the S-ATA connection, they can also be used on the desktop. “Turbo Memory”, on the other hand, is only available for notebooks, and in 2008 the second generation will also reach the desktop. Intel's solution is always tied to a "mainboard chipset" from the same company.
Both concepts require Windows Vista or newer Windows versions, which so far are the only operating systems that can occupy the flash area with the most required data. All other operating systems do not use the flash area.
Newer hybrid hard drives no longer need the operating system to use the flash memory. This process is carried out by a controller in the hard disk itself. This means that such hard drives can be used to the full in any operating system.
Advantages and disadvantages
The theoretical advantages in practice are compared below.
-
Flash usage : HHDs (hybrid hard drives) collect 32MB of data as they are written before the spindle motor starts. The programs that can be started using the special keys on some keyboards are made available again. The other area is available for the most frequently used data.
Instead, “Turbo Memory” is only activated by a driver to be installed later , which is not included in Windows Vista. One half of the flash module then works like that of an HHD, the other half is used like a fast swap memory (see ReadyBoost ). This effectively accelerates PCs with 1 GB RAM to the level of a 2 GB configuration, but cannot be switched off if this is already available. If there is no need to outsource, half of the module remains unused. - Battery life : Both concepts require manual intervention to actually save energy. Since a conventional hard disk is used with “Turbo Memory”, it is shut down by the Windows power options, not by an HHD drive controller. Their default setting provides for a delay of several minutes instead of a second after a hard disk has been accessed. If the setting is corrected to "3 minutes", the battery life is extended by 15%, for example from three to three and a half hours. A comparable effect also occurs with HHDs if the setting "Windows Hybrid hard disk energy saving mode" has been activated in the energy options. ( See also: Green IT )
-
Speed gain : Many benchmarks cannot, in principle, reflect the increased performance of the hybrids - because they use as many, different and large files as possible in order to generate a maximum load. These then exceed the capacity of the buffer many times over. In addition, they are not currently using a recurring access pattern in order to rule out that a drive manufacturer optimizes its product accordingly. However, many of the available performance tests do not do justice to typical notebook use, and HHDs and “Turbo Memory” - similar to a hybrid car under full load - have no advantage in these tests. The former accelerate Windows startup and shutdown by around 20 percent - similar to the start of frequently used programs.
According to initial tests by AnandTech.com, however, “Turbo Memory” does not accelerate the process. The notebook manufacturers Sony and Dell came to the same results and therefore for the time being foregoing this technology. AnandTech examined this together with Intel and actually found the doubling of performance promised by the manufacturer in the “PCMark” test. Outside of the benchmark, however, there were no speed advantages, neither during normal work nor during Windows startup or shutdown.
Even though hybrid hard drives did not come onto the market until 2007, a similar technology was already available more than ten years earlier: The manufacturer Quantum had a SCSI hard drive series called "Rushmore" in its range. This combined a conventional hard drive with SD-RAM the size of a drive instead of flash, which at the time was more of a brake. When the series was discontinued in 2000, these ranged from 130 megabytes to 3.2 gigabytes. All stored data was delivered from the extremely fast "cache" during operation. However, since this was dependent on electricity, the manufacturer equipped the product with batteries to prevent data loss. Its energy allowed the hard drive to start in an emergency and all data to be taken over from the RAM. Because of the high prices for RAM chips, the Rushmore disks were practically unaffordable for private users - they were a thousand times what today's flash chips are. Therefore, the optionally available basic version was no exception: it lacked the relatively cheap hard disk and battery components.
Market situation in 2015
Following Samsung's debut of the first HHD in March 2007, Seagate began manufacturing a model of the same flash size in July. Together with Fujitsu , which has not yet announced an HHD, the named manufacturers founded the “Hybrid Storage Alliance” at the beginning of 2007 in order to better market the advantages of the new technology.
Since Seagate took over Samsung's hard drive division in 2011, Seagate has meanwhile been the only provider of HHDs (mostly referred to as SSHDs by manufacturers) for the end customer market. In the meantime (2014) Seagate offers consumer SSHDs with capacities of 500 to 1000 GB (2.5 inches, generation early 2013) and 1 to 4 TB (3.5 inches, since mid-2013), in which the magnetic storage device is 8 GB MLC flash memory is available as a pure reading cache - twice as much as in the previous generation. Toshiba also presented SSHDs in 2.5-inch format in mid-2013; as with the Seagate SSHDs, the size of the reading cache is 8 GB. In the enterprise segment, Seagate showed SAS-SSHDs with flash cache and up to 600 GB capacity in mid-2013 and advertised them with a performance that is 3 times better than conventional 15,000-rpm hard drives.
At the end of 2013 Western Digital presented a drive called "WD Black² Dual Drive" - a conventional 1 TB magnetic hard disk and 120 GB flash memory in a 2.5 "housing. However, since the flash memory can be addressed separately and is not used as a cache for the hard disk, the WD Black² cannot be called a hybrid hard disk. In order to be able to access both the flash memory and the magnetic hard disk, a special driver is required, without which only the flash memory can be accessed.
With increasing capacity and speed of the flash caches with falling flash prices and at the same time stagnating development in the field of magnetic hard drives, the price and performance of SSHDs have become significantly more attractive compared to conventional HDDs. Many computer manufacturers - above all Lenovo - are now installing SSHDs in portable computers in particular. There they often compete in the high-priced segment with pure SSDs or SSD-HDD combinations (sometimes with chipset caching à la Intel Smart Response, see next paragraph) and in the lower price ranges mostly with pure HDDs.
Intel's solution was introduced in May 2007 with the “Santa Rosa” generation of Centrino . Sony , HP , Dell and MSI refrained from installing the corresponding Intel flash module in their notebooks. After the introduction of Turbo Memory 2.0 (2008), Intel put the Turbo Memory concept on hold. Instead, "Intel Smart Response Technology" was presented in 2011 together with the Z68 chipset, which enables a SATA SSD to be used as a read and write cache for another SATA device. In contrast to "real" SSHDs, the flash cache is managed by a driver at the operating system level. Since the Intel 7 series (mid-2012), the Z, H and Qx7 chipsets of the respective chipset generation have been Smart Response capable.
Other manufacturers such as OCZ (Synapse Cache, end of 2011) and SanDisk (ReadyCache, end of 2012) developed similar concepts in the form of proprietary software, bundled with a small SSD as a cache disk for sale. In view of the sharp drop in the price of SSDs, these concepts have meanwhile lost their importance.
Operating system support for hybrid flash storage
Windows Vista introduced two ways to use flash memory to support conventional hard drives. However, their application only shows performance comparable to SSDs in rare situations, but in return generates little or no additional costs.
For Linux there are special file systems that are adapted to the specifics of raw flash storage , such as JFFS2 , UBIFS and YAFFS ; For SSDs with FTL ( Flash Translation Layer ) and integrated wear leveling, however, just like for USB sticks, conventional file systems such as ext3 are used, sometimes with optimized write access settings (or more suitable file systems such as ZFS , btrfs , NILFS or LogFS ) . Such file systems aim to use flash memory in such a way that its advantages can be maximized. This enables higher speeds and better data integrity control to be achieved.
Windows Vista recognizes the possibilities of HHDs and copies the most frequently used program and operating system files into their flash part. The achievable effects are described above . Vista should also benefit from USB sticks or flash memory cards . It offers to simulate an HHD with them by using part of their storage space as a fast buffer. However, only that which does not fit into the main memory during operation is collected on the flash memory. Representative tests therefore only show a noticeable advantage for the “ ReadyBoost ” idea on PCs with less than one GB of RAM . It thus serves as an easy-to-install RAM extension. Taking into account the prices for main memory, this only makes sense if a correspondingly faster flash memory is already available or an expansion of the main memory is not possible. In contrast to HHDs, the hard drive remains active, which means that neither energy consumption nor volume is reduced. The hard drive also contains an image of the cache that is used when the flash memory is removed. To be on the safe side, the data stored on it is encrypted with 128 bits and the medium is briefly tested for sufficient speed before use. ReadyBoost requires a drive size of 256 Mebibytes, Vista uses a maximum of 4 Gibibytes. The proportion used can be set when connecting. On Linux, a similar method is no longer possible even by the flash memory (English as a swap memory swapping ) is mounted.
Windows XP does not offer either of the two Vista options to use flash memory to increase the speed. With “eBoostr”, the Moscow company MDO Limited offers a tool that implements the “ReadyBoost” idea under XP. It also works with older external flash storage devices; but in order to actually get a gain in speed, the ReadyBoost logo should also be observed here as a guide. This is given to USB sticks and memory cards that achieve a performance level specified by Microsoft. Unlike Vista, the program can also use several flash memories at the same time and display the load distribution between hard disk and flash memory. The target group are PCs that have a USB 2.0 port, but for which RAM expansion is technically or economically not possible.
Pure flash drives
Function and technology
These drives consist of flash and controller chips that are arranged on a circuit board. Their size has no influence on the speed, only on the number of chips. With a small form factor, only lower capacities can still be achieved, but now with high performance. Many models are (partially) clad with plastic or metal to protect the components and - in the second case - to convey solidity. This negates the advantage of the low weight and, in part, the impact resistance due to the intransigence of the metal jacket. The high temperature tolerance, silence and energy efficiency remain.
The so-called NAND chips are used in all price segments in the faster SLC - or in the case of writing, the slower MLC version (see the architecture comparison box). Among the flash technologies, they achieve the best compromise between capacity, price and speed, only their access time is divided into two parts: the operating system and programs start from flash hard disks two to three times as fast as from conventional hard disks, but the disadvantage of the compromise became apparent during writing until 2009 , which could still be concealed in the hybrid concepts - in particular the MLC-based flash SSDs were below the level of normal hard drives with continuous write processes. That has been reversed since then, the fastest hard disks write at least measurably slower than the slowest available flash SSDs. The SSDs were already superior for pure read processes but also for multitasking, i.e. with simultaneous reading and writing. Reading is usually done in a desktop environment, so writing weaknesses are less important here.
As in graphics chips , the speed increases in subsequent product generations are primarily achieved through strong parallelization: For example, some SSDs use a ten-channel controller.
| SLC is to MLC | NAND is to NOR |
|---|---|
| 10 times as durable | 10 times as durable |
| 3 × as fast writing comparable reading |
Writing 4 times as fast, reading 5 times as slowly |
| half as compact | up to 16 × as compact |
| 30% more expensive | 30% cheaper |
| The following technologies should combine the advantages of NAND and NOR: OneNAND (Samsung), mDOC (Sandisk) and ORNAND (Spansion). | |
Only this acceleration compensates for a systemic problem: the internal organization. Flash SSDs are divided into memory blocks . If even one byte is changed in it, the entire block must be rewritten. The drive actually rewrites the blocks internally at the speed specified by the manufacturer. However, users and performance tests only notice the few bytes that have changed. The writing process appears slow. This effect is called write amplification in English . Accordingly, the more the data volume corresponds to the volume of a block, the faster the writing. Files with several megabytes are actually written at the specified transfer rate, because here all bytes in the blocks are changed - the usable rate corresponds to the write rate.
The attempt to reverse this effect at the operating system level did not succeed. The "Managed Flash Technology" from EasyCo. arranged the write commands so that they are as coherent as possible instead of distributed. Regardless of the hardware product, it was too expensive for home use. Instead, the manufacturers added an SDRAM buffer for the controller and carried out a comprehensive optimization of the firmware. Even before the controller itself, this plays the decisive role in the performance of a drive. As before with conventional hard disks, it uses the SDRAM chip as efficiently as possible, but manages the data in it to search for the write area in order to optimize speed and usage distribution . The actual intermediate storage of data usually takes place in the even faster SRAM of the SSD controller chip .
Similar to conventional hard disks, flash drives no longer provide full performance with only a small remaining capacity. With both drive types, the search for the few free storage areas plays the main role. The above-described “write amplification” effect intensifies this on the flash hard drive. Therefore, their manufacturers divert a few percent of the storage capacity for a "spare area" as a performance and wear reserve.
Flash drives are particularly suitable for mobile use, performance-oriented desktops and a few servers. For the latter, some manufacturers offer series with SLC memory in order to cope with the much higher write load. With 10,000 cycles per cell, MLC chips have a tenth of the rewritability of SLC technology.
Market situation
MSystems introduced the first flash-based SSDs back in 1996 . Until 2006, these were reserved for military and other less price-sensitive markets. In March 2006, Samsung then manufactured a model that aimed at a different target market at an eighth of the price: notebooks with 2.5- and 1.8-inch hard drives - and, via an adapter, also desktop PCs. This was an attempt to open up a new market for affordable flash drives. At $ 600 for 32 GB, that didn't work out yet, but Samsung became the market leader with a market share of 45%. So is Apple an important large customers and Mtron SSD in the upper segment controls - with its own controller - also Samsung chips. In 2007, a number of providers with the same objective researched competing products for Samsung's first attempt. In the second quarter of 2008, the representatives of the second generation appeared. In the same year, many manufacturers switched from SLC to MLC memory chips in order to make prices more attractive. With the first products of this type, however, there were more often impractical performance values.
In 2009, products followed with greatly improved performance, which even outperformed SLC drives in some performance points. Since then, price reductions have mostly taken place in parallel with structural downsizing in flash chip production.
Since the flash hard disks are considerably more shock-resistant than conventional hard disks due to the elimination of the sensitive mechanics and a lot more energy-saving, SSDs are particularly suitable for use in mobile computers. From 2012, they were increasingly installed in series in netbooks and notebooks , and later also in stationary computers.
The following table compares the consumer sectors of conventional hard drives, commercial flash SSDs, and industrial flash SSDs.
| conventional hard drive | commercial flash SSD | industrial flash SSD | |
|---|---|---|---|
| Max. capacity | 18 TB | 6.4 TB | 100 TB |
| operating temperatur | 5 to 55 ° C | 0 to 70 ° C | −40 to 85 ° C |
| Write cycles | 10 billion (3 years MTBF ) | from approx. 1,000 (TLC) / approx. 3,000 (MLC) up to approx. 100,000 (SLC) / flash cell | 1–5 million / flash cell ("hand-picked" chips) |
| Data retention | not specified | ten years | ten years |
| Flight recorder -suitable | No | No | Yes |
| Secure deletion | Yes | partially | Yes |
| SMART | Yes | partially | Yes |
Problems
The quality of the built-in NAND flash varies . In addition to NAND flash classified as “Class 1”, NAND flash classified as “Class 2” is also installed in SSDs. The SSD manufacturer OWC even found “off-spec” NAND in the SSD in a sample of SSDs from its competitor OCZ - components whose properties are not within the specification . These are chips that actually did not pass quality assurance for use in SSDs according to the NAND manufacturer. Other manufacturers, such as Samsung in the 840 SSD series, also rely on TLC NAND memory cells. Compared to SLC ( single-level cell ), TLC ( triple-level cell , three-level memory cells , but de facto 3 bits) has additional voltage levels, which means that even more data can be stored per memory cell. Due to the smaller intervals between these levels and the resulting difficulty in always reading these levels correctly, the service life of the memory cells decreases as the number of bits increases. Manufacturers are trying to counteract this by adapting production technology (e.g. 3D-V-NAND technology).
comparison
The following table compares the properties of common computer storage methods of the mass market in 2016, i.e. not the high-end sector in the server area . As a rule, maximum values are given. The speeds in particular can be significantly lower depending on the model.
| MLC NAND Flash Drive 1.0 ″ to 3.5 ″ |
RAM disk as part of the main memory | Hard disk 1.0 to 3.5 ″ |
|
|---|---|---|---|
| Size (no raid drives) | Up to 16 TB | up to 32 GB per module | up to 14 TB |
| connection | IDE / (P) ATA , SATA , mSATA , PCIe , M.2 | mainly DIMM connector | SCSI , IDE / (P) ATA, SATA, SAS |
| Read (no RAID ) | up to 510 MB / s | up to 51,200 MB / s | up to approx. 227 MB / s |
| Write (no RAID) | up to 490 MB / s | up to 51,200 MB / s | up to approx. 160 MB / s |
| Read average access time | from 0.031 ms | 0.000.02 ms | from 3.5 ms |
| Write average access time | from 0.023 ms | 0.000.02 ms | from 3.5 ms |
| Overwritable (cycles) | 3 to 10 thousand times (MLC) | > 10 15 | approx. 10 billion (3 years) |
| Can be stored at | −45–85 ° C | −25-85 ° C | −40-70 ° C |
| Shock resistance - operation | 1,500 g | approx. 1,000 g (soldered vibration-proof) | 60 g |
| Shock resistance - storage | 1,500 g | approx. 1,000 g (similar to SSD) | 350 g |
| Consumption - rest | 0.1-1.3 W. | 1 W per SDRAM module | 4 W and higher |
| Consumption - access | 0.5-5.8 W | 8 W per SDRAM module | 6 W and higher |
| Shutdown behavior | problem-free | Loss of data if no backup is made on SSD / hard drive | problem-free |
| Behavior in the event of a power failure | problem-free with a backup capacitor, otherwise data loss is possible | Data loss | Loss of data possible |
| Silent | Noises possible if there is a defect | Yes | No |
| Remarks | mostly support SMART | Size limited by motherboard or adapter required, not bootable | support SMART |
Flash specifics
Wear and failure prediction (SMART)
The lifespan of an SSD should be specified in one of the following two forms on its data sheet:
- either in n DW / D ( n drive writes per day , means “n complete drive overwrites per day”) + indication of the period for which the SSD can withstand this constant load;
- or in m TBW ( m terabytes written (also: total bytes written ), TBW for short , means “total terabytes written”). (usually also a maximum guarantee period)
For example, a Samsung 970 EVO M.2 SSD (2018) with 1 TB has a lifespan of 600 TBW (or 5 years; whichever comes first).
Conventional and flash hard drives wear out over time. While the former results from the wear and tear of the plate surfaces and the mechanics, an electrical effect has a limiting effect in the case of flash technology . Read processes from intact cells are unlimited, but depending on the quality, a flash cell can only complete between 3,000 (MLC 25 nm) and 100,000 (SLC) write processes. Then it can forget its content, but above all the cells cannot be erased after the end of their life. Flash memory would be defective after just a few days. The effect wear leveling algorithms contrary. The controller in the flash drive distributes the write operations to all memory cells so that each is written as often as possible. The algorithms used for this are manufacturer-specific, but in any case neither visible nor controllable from the rest of the computer. This distribution process is available in different stages. For example, a flash drive often uses more complex controllers than a USB stick and very few removable storage media do not use any. Software solutions such as those in Windows Vista or file systems such as JFFS2 or the Flash-Friendly File System under Linux can help out here.
Depending on the expansion stage, the process leads to a durability that comes close to or exceeds that of conventional hard drives. A side effect of all distribution methods, however, is that secure deletion is no longer possible. The background to this is described in the sections Secure Erase and Defragmentation .
A failure prediction by SMART , as with conventional hard drives, is also included in almost all Flash SSDs, but not possible with all programs. The situation is justified by the standard itself. It only includes communication with the drive in order to read out the SMART values. Their meaning and scaling are not specified. For conventional hard drives, however, a de facto standard has emerged over the years, which is missing for flash SSDs.
The test focuses on the number of erase cycles performed and the existence of sufficient reserve blocks. If the intended limits are exceeded here, the drive goes into read-only mode to be on the safe side. Since with good wear-leveling procedures all normal sectors are worn out at a similar time, a failure after the use of the first reserve sectors is probably close.
Other methods used to increase the service life include avoiding unnecessary writing. This includes Native Command Queuing (NCQ), which has the effect of writing the same multiple data block in the cache only in its most recent version and discarding the obsolete copies from the cache. Data compression of the data to be written in the controller of the SSD means that individual blocks do not have to be written. This does not work with data that has already been compressed, as the amount of data cannot be reduced any further. If further data blocks are occupied by content that has failed for a longer period of time, these will be marked as unusable with SMART, unless otherwise managed.
Methods of distribution of use
Files are always written as a bit sequence. MLC -Flashzellen typically include two bits and are called pages or blocks (English pages combined with 4096-byte size). Whole memory blocks are always addressed by the control unit. When reading individually, when writing, they are again combined to form an erasable block . This contains 64 or 128 blocks and is therefore 256 or 512 KiB in size. Each time a change is made to one of its blocks, it is initially not deleted, but is initially marked as out of date . The next free block of the same erasable block is used for writing . Only when all the blocks of an erasable block are no longer up-to-date is it completely deleted. Thus, for each changed byte, the previous blocks with the changes to be made must be copied to the next. This allows the data to be written to increase from a few changed bytes to several kilobytes. This multiplication is therefore also referred to as write amplification . This would result in an unacceptable shelf life. In the following example, a text file is revised and saved four times.
| Writing process | 1 | 2 | 3 | 4th | continue as 2 | |||
|---|---|---|---|---|---|---|---|---|
| block | ||||||||
| 1 | File.txt | out of date | out of date | Clear | File.txt | ... | ||
| 2 | empty | File.txt | out of date | Clear | empty | ... | ||
| 3 | empty | empty | File.txt | Clear | empty | ... | ||
| 4th | empty | empty | empty | empty | empty | ... | ||
| 5 | ... | ... | ... | ... | ... | ... | ||
| Legend: | ||||||||
| An " erasable " block: for the sake of clarity, only includes three memory blocks. Each block is 2 or 4 kibibytes. | ||||||||
- Dynamic wear leveling
- If an erasable block is to be written on, the least worn block is selected from those not yet used. This is comparatively easy to implement in the controller, but has the disadvantage that if the drive is well filled, the little free space will be worn out faster. The write cycles increase by a factor of 25 compared to a lack of wear leveling.
- Static Wear Leveling
- If a block is to be written on, the one that is least worn is selected here. If this is already occupied, its data is relocated to another, then the new data is written. This requires a slightly more complex controller, but results in very even wear. The write cycles increase by a factor of 100 compared to a lack of wear leveling.
- Broken blocks
- If a write attempt on a block fails, it is marked as unusable as with conventional hard disks and a reserve block is activated.
Swapping memory on flash SSDs
The best way to assess the suitability of flash drives for accommodating the swap memory (or swap file ) of an operating system is to analyze the accesses to this memory. Microsoft did one while working on Windows 7 . The evaluation revealed an access pattern of short, distributed reading and longer, coherent writing. This corresponds to the strengths of flash storage. Read accesses exceeded writes by forty times, while about two thirds of the read accesses were up to 4 KB in size and about two thirds of the write accesses were at least 128 KB in length. Since this corresponds roughly to an erasable block , according to Microsoft there are hardly any more suitable applications for flash drives than swap memory. However, fast SSDs should be preferred for this.
Secure deletion
Ordinary operating systems do not delete the file contents themselves, but only remove the entry in the table of contents of the file system. This speeds up the deletion process, but also enables the file to be restored. To prevent this (and thus espionage), there are programs that save the files, i. H. actually delete their contents. For this purpose, these programs instruct to overwrite all sectors belonging to the file multiple times - with random data if necessary. Mechanical hard drives can thus be securely erased.
However, the sectors reported to the outside by flash memories no longer have anything to do with the actual storage locations. This is due to their distribution of use , which directs write operations to the previously least used blocks, which at most just happen to be the ones in which the file is located. Their content would thus remain, while the overwrite attempts would be saved elsewhere. Externally, the sectors addressed by the program remain consistent: If you read them out, you get the new data. The redistribution happens imperceptibly to the operating system and programs running on it in the SSD controller chip. It takes place all the more, the more sectors that have not been written to or have been released by TRIM after the last formatting are present on the drive - a well-filled drive is, from this point of view, a "safer" one.
In order to use this security leak and to be able to access the file that has not been physically deleted, firmware would have to be programmed and installed that can read out all blocks. With their installation, however, the information on the previous usage distribution would probably be lost. So there was a lack of knowledge about which blocks belong to a file that had apparently been deleted by overwriting and in which order. Cryptography manufacturers still warn against the use of SSDs, as at least keys could be found.
The problem can only be remedied by a controller who, if desired, can temporarily switch off the distribution of use and thus enable secure erasure ("Secure Erase"). Corresponding drives can only be found in the high-price segment, for example from M-Systems or ATP. These then also contain deletion algorithms based on the US Air Force or Navy standard.
There is no complete deletion option for home use. This is due to the non-addressable reserve memory ("spare area") of the SSDs, which is only accessible to the controller. This area serves both as a resting place for the most worn sectors, as well as to increase speed . Modern drives offer secure erasure that resets the SSD to the delivery state and also overwrites reserve blocks.
If an SSD reaches the end of its life, it can no longer be written to, even if only in part, and it can no longer be completely erased. Accordingly, all that remains is physical destruction to finally destroy the data.
Defragmentation (Windows)
Defragmentation is not necessary due to the functional principle: memory blocks are addressed, data is not placed and accessed electromechanically as with conventional hard drives (HDD). A sequential arrangement of the data, as generated by defragmentation, is therefore unnecessary with an SSD. All SSDs also have internal algorithms that ensure that all cells are worn out equally; the data is placed exclusively by the SSD controller; a defragmentation program has no influence whatsoever. All SSD manufacturers advise against defragmenting. Since SSDs should not be defragmented by their technology, since constant writing even reduces the lifespan of the disks, Microsoft recommends switching off all software cache systems such as Prefetch and SuperFetch in addition to switching off defragmentation . It is recommended to turn off automatic defragmentation programs; There are instructions on how to do this, especially under the Windows NTFS file system.
All common operating systems (e.g. Windows version 7 or higher) recognize when an SSD is installed and automatically deactivate all specific HDD features. However, it is recommended to check all settings.
Loss of performance when used (TRIM and Garbage Collection)
background
The file system only removes "deleted" files from the table of contents, but the actual file remains stored. This allows it to be restored, and large amounts of data can also be "deleted" very quickly. The next time you write to an area released in this way, the previous content must therefore first be deleted. After some time of use, every area of the drive is occupied with either current or not yet actually deleted content. This was not a problem with hard disks, since their magnetization states can merge directly into one another. (For them, actually deleting the files would have been a waste of resources.) Flash memory, on the other hand, first has to empty the flash cells that are still full before the new file can be written to them in the second pass. This double work can be traced back to the typing time that is up to twice as long. However, this only affects distributed, short write processes that are smaller than erasable blocks - because by filling with current and not yet deleted data, their individual blocks are filled, which means that the entire "erasable block" has to be rewritten with every change - including the actually "deleted" file fragments. Even in these cases the speeds often remain above the - constant - level of conventional hard drives.
activities
To remedy this situation, from mid-2009 SSDs can delete the exposed areas before they can be used again. This takes place on the one hand by logic in the drive ( garbage collection ) and on the other hand can be controlled by the operating system (via the TRIM command). The former requires only the reaction in the drive firmware, the latter requires the support of a new ATA command from host operating system of the receiving drive and - if present - on the forwarding RAID - Controller .
But both procedures involve neither the About write a file still the case almost filled drives (because here there is little free areas) but allow only ready for writing prepare the exempted areas. However, the preparation does not correspond to an immediate deletion when emptying the recycle bin. This takes place at an unpredictable point in time, at the latest when the area is next written. It is deleted and written to directly, the prior reading is saved: This results in the speed advantage.
The lasting advantage of TRIM lies in the more effective avoidance of rewriting files that have already been deleted from the recycle bin. That protects the flash cells. The support of TRIM by the installed drive is visible, for example, by the CrystalDiskInfo tool in the "Supported Features" line; the Firmware line shows the installed firmware version.
| Controller | Garbage collection | TRIM support |
|---|---|---|
| Indilinx Barefoot and ECO | from version 1916 | from version 1819 |
| Intel X-25M G1 | always included | not available |
| Intel X-25M G2 | always included | from version 02HD |
| Samsung RBB | from 1901Q / 19C1Q | from 1901Q / 19C1Q |
| SandForce SF-1x00 / 2x00 | always included | always included |
| Toshiba Daikoku 2 | always included | always included |
The way garbage collection works is not published by the manufacturers, but it probably works through the use of the spare sectors in a drive. (The over-provisioning factor indicates this number; if it is 1.1, the drive has 10 percent of it. This area is also known as the “spare area”.) The drive itself cannot know which sectors are available for overwriting Contain data. However, this is what it learns when sectors are to be overwritten: their previous content is obviously optional. Therefore, it redirects the new data to the empty reserve sectors. That is fast, and the controller now knows that the sectors originally controlled are really no longer needed. It now deletes these while idling, which frees them. This procedure does not correct the cause of the problem, but it does largely correct the effects. However, the drive needs idle time to erase the sectors.
Drives without automatic garbage collection or TRIM support can only be reset to the factory settings using programs such as Secure Erase. In doing so, the file system is deleted, which also releases all the blocks that it has released - but which still contain old files.
Loss of performance on Windows
Further performance losses may occur if Windows does not automatically deactivate defragmentation , prefetcher and superfetch after installing the SSD. This effect could arise when copying an existing installation from hard disk to SSD. For the full performance of an SSD, you should therefore deactivate these functions manually if necessary.
Prefetch only has a noticeable advantage for storage media with relatively high access times, such as HDDs. The original problem that the prefetcher was supposed to fix is calling different segments of the same file at different times. The prefetcher records the boot process of the operating system and the start process of programs. Based on this information, the operating system creates trace files so that files can be accessed more efficiently, for example required segments of a file can be read in one piece. With SSDs, this speed advantage is hardly noticeable thanks to the short access times. However, when prefetching, a lot of data (trace files) is saved and updated if necessary. This results in unnecessary write access and thus a shortened service life of the SSD.
The advantage of SuperFetch (ReadyBoost), in which frequently used data is loaded into the RAM based on empirical values, before it is needed, is also eliminated. The access times and data transfer rates of SSDs are so short that it makes no noticeable difference compared to conventional RAM, it can only be measured in the range of microseconds. However, the SuperFetch is also not harmful to the SSD or the memory load, as the area reserved for it in the RAM is immediately discarded and released as soon as an application needs it.
Windows version 7 or higher recognizes during installation what type of data carrier it is. Windows switches off both SuperFetch and Prefetch on SSD data carriers. With parallel use of SSD and HDD, Windows only deactivates these functions for the SSD and not for the magnetic hard disk (HDD) as well.
Alignment
The smallest recordable unit of a flash memory is the memory page (also English page called). The smallest writable unit in a file system is called an allocation unit (or cluster ). When formatting, this should ideally be set to the size of a page. If the size of an allocation unit exceeds that of the memory page, several pages would otherwise have to be rewritten unnecessarily with every change.
Alignment of partitions is a second step towards the optimal data structure . If a partition begins in the middle of an erasable block , the allocation units of the file system also shift; some then extend over a block boundary. For every change in these units, both blocks are written and therefore deleted more often. However, the effect of this additional effort is very different depending on the product and ranges from hardly measurable to one third of the write performance in the case of random access. All sequential write and all read accesses are not affected by the "alignment".
In order to align a partition to the boundaries of the erasable blocks, the block size would have to be found out. However, software cannot read these out. However, all MLC drives that have been in use since 2009 consistently use a size of 512 kibibytes. Manual re-measurement is therefore rarely necessary.
The specification of the "alignment" in the AS-SSD benchmark is carried out using the allocation unit used. If the program sees the "alignment" as "OK", it does not mean that the partition is aligned directly with the boundary of the erasable block, but that no allocation unit of the file system is located in two blocks at the same time. Since the first erase block always starts at 0 bytes, it is sufficient if the starting position of each partition is divisible by the size of the allocation units (typically 4 KiB ).
If in doubt, a larger value can be used directly. Following this, Windows Vista and 7 use an alignment to 1 Mebibyte . This value can be divided by all current erasable block sizes without a remainder and therefore causes a correct alignment for each SSD. Linux users need to consider several factors for alignment. Users of previous Windows versions can either set up a prepared partition with the Vista and 7 installation / recovery CDs or create them themselves using the on-board tools. Without a manual procedure, the older Windows editions start the partitions at 31.5 Kibibyte and are therefore not aligned with the common erasable blocks .
The "offset" of existing partitions can be seen in Windows on the command prompt with the Diskpart command list partitionafter selecting the SSD select disk <Nummer>. The number can be determined using the data carrier size using list disk. The information here is inaccurate, however, as it is rounded to whole kilobytes. It is therefore better select partition <Nummer>to detail partitioncall up first and then . Alternatively, the same information can also be found msinfo32under System overview \ Components \ Storage \ Data carrier as a partition start offset.
literature
- Martin Bayer: Everything you always wanted to know about Solid State Drives. In: Computerwoche , June 24, 2008.
- Daniel Menzel: Performanter Storage in 2019 In: Informatik Aktuell (magazine) , December 30, 2019.
Web links
- Sebastian Geertz, Axel Wilkening: Development potential of solid state drives. Winfwiki, June 14, 2009 (case study from the University of Applied Sciences for Economics & Management (FOM) Hamburg / Essen on the subject of "Solid State Drives")
- Lee Hutchinson: Solid-state revolution: in-depth on how SSDs really work. Ars Technica, July 4th 2012.
Individual evidence
- ↑ Data ExaDrive DC100 from Nimbus is the world's largest SSD with 100 TB. Pcwelt.de , March 20, 2018, accessed March 20, 2018 .
- ↑ Samsung SSD 840 (250GB) Review. AnandTech, October 8, 2012, accessed December 30, 2019 .
- ↑ [1]
- ^ Vittorio Ferretti: Dictionary of electronics, data technology, telecommunications and media: German-English . Birkhäuser, 2004, ISBN 978-3-540-40693-8 , pp. 367 .
- ↑ mass storage: prices, security, virtualization; Expensive, expensive, solid state memory - chip , 14 May 2010
- ↑ Price comparison (€ / GB) SSDs -1TB. geizhals.at, October 25, 2014, accessed October 25, 2014 .
- ↑ Price comparison (€ / GB) hard drives 500GB-8TB. geizhals.at, October 25, 2014, accessed October 25, 2014 .
- ↑ Marvell & SandForce: Firmware updates for SSDs. hardware-infos.com, May 20, 2010, accessed February 3, 2012 .
- ↑ Quick troubleshooting - firmware update for Crucials M4. computerbase.de, January 14, 2012, accessed on April 14, 2012 .
- ↑ Intel's SSD 320 problem still exists despite the firmware update. ht4u.de, August 22, 2011, accessed April 14, 2012 .
- ↑ Thomas M. Rent: Origin of Solid State Drives. storagereview.com, March 20, 2010, accessed October 25, 2014 .
- ↑ Helmut Weber: Microprogramming the IBM System / 36O Model September 30 , 1967, accessed on October 25, 2014 .
- ↑ Auxiliary memory . In: Encyclopædia Britannica .
- ↑ IBM User's Guide, Thirteenth Edition. (No longer available online.) Web.utk.edu, June 30, 1960, archived from the original on February 14, 2015 ; Retrieved October 25, 2014 . Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice.
- ^ Zsolt Kerekes: Charting the 30 Year Rise of the Solid State Disk Market. storagesearch.com, accessed October 25, 2014 .
- ↑ Dataram Corp: 1977 Annual Report. (PDF) Retrieved October 25, 2014 .
- ↑ SSD Market History - Charting the 30 Year Rise of the Solid State Disk Market. storagesearch.com, accessed October 25, 2014 .
- ^ Fred Moore: Enterprise Storage Report for the 1990s. (PDF) Storage Technology Corporation, accessed October 25, 2014 .
- ↑ David H. Ahl: The Sharp PC-5000; a desktop computer in a portable package . In: Creative Computing . tape 10 , no. January 1 , 1984.
- ↑ Patrick Waurzyniak: Battery-Powered Mass Storage System Offered . InfoWorld Media Group, September 8, 1986, ISSN 0199-6649 , p. 54 ff .
- ^ Zsolt Kerekes: SSD Market History - Charting the 30 Year Rise of the Solid State Disk Market. storagesearch.com, accessed October 25, 2014 .
- ^ EMC Corporation. Archived from the original on June 21, 2000 ; Retrieved October 25, 2014 .
- ↑ Psion MC 400 Mobile Computer. retrocosm.net, March 29, 2012, accessed October 25, 2014 .
- ↑ Odagiri Hiroyuki, Goto Akira, Sunami Atsushi, Nelson Richard R .: Intellectual Property Rights, Development, and Catch Up: An International Comparative Study . Oxford University Press, 2010, ISBN 0-19-957475-8 , pp. 224-227 .
- ↑ Cash Kelly: Flash SSDs - Inferior Technology or Closet Superstar? BiTMICRO, archived from the original on July 19, 2011 ; Retrieved October 25, 2014 .
- ^ Gary Drossel: Solid-state drives meet military storage security requirements. (PDF) Military Embedded Systems, February 2007, accessed October 25, 2014 .
- ↑ BiTMICRO 1999 News Releases. BiTMICRO, 1999, archived from the original on September 16, 2009 ; Retrieved October 25, 2014 .
- ↑ Fusion-io announces ioDrive, placing the power of a SAN in the palm of your hand. PRNewswire, September 25, 2007, accessed October 25, 2014 .
- ↑ OCZ's New Blazing Fast 1TB Z SSD Drive. Tom's Hardware, March 4, 2009, accessed October 25, 2014 .
- ↑ Ng Jansen: Micron Announces World's First Native 6Gbps SATA Solid State Drive. (No longer available online.) DailyTech, December 2, 2009, archived from the original on December 5, 2009 ; Retrieved October 25, 2014 . Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice.
- ↑ Chris Mellor: EMC has changed enterprise disk storage for ever: First into the enterprise flash breech. Techworld, accessed October 25, 2014 .
- ↑ Barry A. Burke: 1.040: efd - what's in a name? The Storage Anarchist, February 18, 2009; archived from the original on May 1, 2015 ; accessed on February 15, 2016 .
- ↑ Anand Lal Shimpi: The Intel SSD DC S3700 (200GB) Review? AnandTech, November 9, 2012, accessed October 25, 2014 .
- ^ StorageTek Forum 2004 conference guide. (PDF; 797 kB) Thirty-Five Years of Innovation. (No longer available online.) StorageTek, Oct. 2004, p. 6 , archived from the original on Jan. 5, 2012 ; accessed on October 17, 2010 (English). Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice.
- ↑ fusion io performance compared to normal SLC flash SSDs
- ↑ fusion-io SSD data sheet. see "ioDrive Octal Capacity". INCOM Storage, accessed June 24, 2013 .
- ↑ DTS "Platinum HDD"
- ↑ Here and here you can see the Winbond W9751G6JB-25 chip on the board of the Seagate ST3000DM001, which is a DDR2 SDRAM chip: data sheet ( Memento from March 5, 2014 in the Internet Archive ) (on the meaning of - 25 at the end see page 4; PDF; 1.5 MB)
- ↑ Intel "Turbo Memory"
- ↑ Does “Turbo Memory” really accelerate?
- ↑ HHD series Samsung ( Memento of the original from June 26, 2007 in the Internet Archive ) Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice.
- ↑ Seagate HHD series
- ↑ Hybrid Storage Alliance ( Memento of December 7, 2008 in the Internet Archive )
- ↑ Seagate SSHD specifications.
- ↑ News on Golem.de about new SSHDs from Toshiba.
- ↑ Seagate Press Release Enterprise SSHDs ( Memento July 28, 2013 in the Internet Archive )
- ↑ FAQ on the WD Black Dual
- ↑ Market overview notebooks with SSHD on Geizhals.de, accessed on June 20, 2014
- ↑ Intel "Turbo Memory" 2.0
- ↑ Presentation of the Z68 chipset with Smart Response.
- ↑ Intel product page on Smart Response Technology.
- ↑ OCZ Synapse Cache product page ( Memento from March 24, 2013 in the Internet Archive )
- ↑ SanDisk ReadyCache product page. ( Memento of the original from July 14, 2014 in the Internet Archive ) Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice.
- ↑ Price development of the Samsung 840 Pro on Geizhals.de as an example of the drop in prices for SSDs, accessed on June 20, 2014
- ↑ What does “ReadyBoost” bring? Computerbase.de
- ↑ eBoostr - ReadyBoost alternative for Windows XP.
- ↑ a b FC test - continued
- ↑ Real write speed of faster hard drives and SSDs from 2010.
- ↑ Anandtech: Multitasking
- ↑ Inside the X25-M SSD: 10-channel flash ( page no longer available , search in web archives ) Info: The link was automatically marked as defective. Please check the link according to the instructions and then remove this notice.
- ↑ Patent DE10349595 : Method for writing memory sectors in a memory that can be erased in blocks.
- ↑ Patent EP1514171 : Method for restoring administrative data records from a memory that can be erased in blocks.
- ↑ a b Patent DE102005001038 : Method for converting logical to real block addresses in flash memories.
- ↑ SSD cache design
- ↑ Too low-performance first MLC drives
- ↑ Overview of prices for current flash SSDs in the ComputerBase forum computerbase.de
- ↑ An overview of the models available for private individuals and their prices and price developments can be found in the heise price comparison
- ↑ The next round of price reductions for Flash SSDs with a switch to 25 nm is expected.
- ↑ https://www.heise.de/news/WD-Gold-18-TByte-Bisher-groesste-Festplatte-vorbestellbar-4793812.html
- ↑ 6.4 TB SSD hard drive , in idealo.de
- ↑ Michael Günsch: ExaDrive DC100: The first 100 TB SSD comes from Nimbus Data. Retrieved March 17, 2020 .
- ↑ Write cycles SLC and MLC - paragraph How long have you got before the disk is trashed?
- ↑ Not All SSDs Are Created Equal: The Story Continues. owc blog, March 18, 2011, accessed April 14, 2012 .
- ↑ a b AS-SSD Incompressible Sequential Performance (Samsung SSD 840 Pro (256 GB) Review).
- ↑ a b DDR3-1600 SDRAM in quad-channel.
- ↑ a b computerbase.de
- ↑ a b test: Samsung 840 EVO 1 TB (MZ-7TE1T0BW) (SSD). chip.de
- ↑ Quantum Dot Storage. ( Memento of the original from March 4, 2016 in the Internet Archive ) Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice. (PDF) sol.physik.tu-berlin.de, p. 6
- ↑ Long-term archiving # Shelf life of the carrier media hard disks during operation.
- ↑ The plug connections only withstand up to 30 g ; Technical specification of 184-pin DIMM sockets. (PDF; 90 kB), Technical specification of SATA connectors. (PDF; 167 kB)
- ↑ Anand Lal Shimpi: Power Consumption. AnandTech, November 23, 2011 (article).
- ↑ As a matter of principle, a backup capacitor in the SSD cannot completely prevent data loss due to a power failure, since a consistent state does not necessarily arise after the write cache of the SSD has been emptied (e.g. if the data to be written is transferred from the main memory to the cache of the SSD the interface is not yet completed).
- ↑ http://www.tomshardware.de/ssd-mythos-flash-speicher-festplatten-hdd,testberichte-241627-4.html
- ↑ http://www.tecchannel.de/a/und-sie-laermen-doch-pfeiftoene-bei-ssds,2026993
- ↑ 'TBW' = 'total bytes written': samsung.com: 970 PRO
- ↑ Tech Brief - Matching SSD Endurance to Common Enterprise Applications . Retrieved June 13, 2020.
- ↑ Product: Samsung 970 EVO NVMe M.2 SSD 1TB . Retrieved June 13, 2020.
- ↑ Procedure for distribution of use. ( Memento of the original from March 5, 2016 in the Internet Archive ) Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice. Format: PDF, 0.8 MB
- ↑ Tom Schäfer: SSD hard disk lifespan: Does the SSD really last longer? August 26, 2013, archived from the original on January 31, 2015 ; accessed on February 15, 2016 .
- ↑ Test for the durability of flash cells. ( Memento from November 17, 2006 in the Internet Archive ) In: c't (paragraph half-life)
- ↑ Last draft of the currently valid standard, the SMART attributes are still missing ( Memento of December 11, 2009 in the Internet Archive )
- ↑ SMART functions of an SSD exemplary. ( Memento of the original from August 28, 2008 in the Internet Archive ) Info: The archive link was automatically inserted and not yet checked. Please check the original and archive link according to the instructions and then remove this notice. (P. 31) Format: PDF, 0.5 MB.
- ↑ The SandForce controller in detail. Retrieved November 29, 2013
- ↑ Tarun Iyer: LSI Showcases Next-Generation SandForce SSD Controllers. tomshardware.com from August 26, 2013
- ↑ DuraWrite TM Virtual Capacity ( Memento December 3, 2013 in the Internet Archive ) on lsi.com, accessed November 29, 2013
- ↑ The SSD Anthology: Understanding SSDs and New Drives from OCZ. AnandTech, March 18, 2009 (English) accessed August 1, 2012
- ↑ Suitability of Flash SSDs for the swap file.
- ↑ The reserve area of an SSD is memory that cannot be overwritten.
- ↑ Windows 7 & SSD: Defragmentation, Superfetch, Prefetch ( Memento from September 27, 2013 in the Internet Archive ) (English), archived from the original ( Memento from September 27, 2013 in the Internet Archive ) on September 27, 2013
- ↑ SSD and SuperFetch (Windows 7) In: c't , 6/2012; Retrieved January 3, 2015
- ↑ Windows and SSD - Configuration and Optimization ( Memento of March 12, 2009 in the Internet Archive ) (Windows 7 and 8) Microsoft Community Wiki, accessed on January 3, 2015
- ↑ anandtech.com
- ↑ anandtech.com
- ↑ Show trim support.
- ↑ Firmware Indilinx changelog.
- ↑ Firmware Samsung ( Memento from March 12, 2009 in the Internet Archive )
- ↑ SandForce Recycler for Garbage Collection.
- ↑ Toshiba Daikoku 2 Review of Garbage Collection and TRIM.
- ↑ Effect of garbage collection.
- ↑ Secure Erase to restore the factory performance in the event of no garbage collection and data loss ( Memento from July 2, 2012 in the Internet Archive )
- ↑ Effect of unaligned partitions that are not aligned with erasable blocks.
- ↑ Manual determination of the erasable block size.
- ↑ AS SSD benchmark
- ↑ Linux and the alignment of partitions to erasable blocks.
- ↑ Alignment: optimize partitions. - Article (page 5 of 8) at Chip online , from September 4, 2009 (accessed on: August 1, 2012)
