Flash Memory

Flash memory has come a long way since its first beginnings. Back in the early days of computing, flash memory was quite the exotic offering, reserved for specialist applications that could justify the high cost of this type of storage.

In the business domain, among the first flash storage devices were EPROMs. Once an application had been debugged and thoroughly tested, it could be written or ‘flashed’ onto a silicon stratum and then employed in larger systems, such as mainframes or supercomputers.

The superior speed of on-chip storage (read-only, or expensive read-write, as it emerged into the market) surpassed that of the standard magnetic storage devices of the day.


The term "flash" was coined by the technology's inventor Fujio Masuoka, an employee of Toshiba at the time, in 1980. He used the term as the process of writing data in this new, experimental manner reminded a colleague of the flash of a camera.

As is often the case in technology, the speed at which flash memory evolved was quick. Before long, increases in the efficiency of manufacturing processes meant that memory capacities grew quickly, alongside reliability levels, error correction abilities and manufacturing accuracy.

Microprocessors at the heart of computers get faster for a number of reasons, but it is due in part to miniaturisation. The same is true of flash memory. As memory cells become smaller (down to 12nm at present) more capacity is available per square inch; this is termed areal density in both the flash and magnetic media industries.

But in order for Moore's law [https://en.wikipedia.org/wiki/Moore%27s_law]of microprocessor power to be reflected in flash memory, miniaturisation is only one way to increase the capacity and reliability of flash memory. In 2007 Toshiba announced 3D-NAND technology[http://ieeexplore.ieee.org/document/4547607/], which the company calls BiCs.

The 3D NAND or V-NAND (vertical NAND) method of storage uses layers of memory strata which can be written and read more quickly than a single layer, roughly to a factor of the number of layers. Currently, 64 layers are commonplace, although Western Digital offers a 96 layer 3D-NAND memory.

The word "roughly" appears in the paragraph above due (amongst other issues) to the need for error correction, which manifests in several[https://en.wikipedia.org/wiki
eed%E2%80%93Solomon_error_correction] forms[https://en.wikipedia.org/wiki/BCH_code]. Error correction tech is of course present in any storage medium.

Error correction is of particular pertinence to flash memory, as it compensates for one of NAND memory's particular characteristics, which is a finite shelf-life, as defined by read-write counts.

In short, flash memory cells deteriorate each time they are written, or programmed, to use the vernacular. Currently, flash memory (which appears, for instance in SSD drives, now becoming pretty much standard in some settings) can be written to between 1,000 and 1,000,000 times, before it becomes effectively unusable.

However, as the technology improves, the standard life of flash SSD devices, to take one example, is rapidly approaching that of HDD drives. The total life of flash devices is dependent on several factors, one being whether they comprise of single or multiple layers of memory, manufacturer, node size, and so forth.

Of course, NAND memory is used in a multitude of settings right across the technology gamut, with hard drives and memory cards of whatever flavour being but two common aspects of flash memory use.

The latest offerings to come to market include micro SD cards which are accredited as "A1" by the SD Association card to their 5.1 specifications, making them highly reliable.

A further instance is the embedding of wireless tech onto the flash device, which allows it to communicate, or at least be accessible by any device on a LAN. (The possibilities of this go beyond the ability to capture and preserve definitive footage of mythical beasts - see below.)

Like many technologies, flash has its pros and cons:

- rapid read and write speeds
- smaller in size than traditional, magnetic media
- comparatively lower power requirements
- generate less heat
- highly effective in ACID transactions, such as in databases

- limited read/write frequency
- relatively high cost
- potentially limited at-rest life (around one year at room temperature)

While there have been predictions about supply outstripping demand for flash storage in 2018 and beyond, it seems that current demand for the benefits of flash are ensuring that there will be no significant drop in price, especially towards Q4 2018, when demand is thought to be particularly strong.

Continuing strong demand for flash is fuelled in part due to the enterprise's demand for data storage growing, alongside its demands for more processing power. Demand is being driven by many factors, the most notable being the need for large data lakes, IoT data capture, information capture for machine-learning applications and, of course, consumers' desire to take more pictures, audio, and video.

[Embed this video: https://www.youtube.com/watch?v=QqRcTibAZuI]


Blog entry is closed for new comments.