10 years of Kryder’s Law: how flash will shape the next 10 years
Wed 9 Dec 2015
Flash storage is now ubiquitous in the enterprise. Introduced to the IT market over thirty years ago, it has become the go-to storage of choice over the past decade. It’s everywhere, from the world of consumer computing, built into hand held devices such as smartphones, laptops and tablets as well as memory cards for your digital camera, to the business world, inside every company’s data centre.
The prevalence of flash technology has come about because, while it offers read and write speeds hundreds of times faster than traditional HDDs, it has also recently decreased in price to a level that makes it accessible to everyone. According to industry experts, flash prices have decreased by three quarters in the 18 months prior to July 2015, and will continue to do so, due to its increasing popularity. There are more and more use cases for flash emerging every day, meaning manufacturing prices have decreased due to economies of scale.
As flash storage is becoming cheaper as time goes on, the capacity is also increasing while the physical size decreases. The smaller physical architecture of flash storage and the associated efficiency improvements that go with this mean that there are also changes in the form of technology, which is leading to developments in deployed technologies, including faster processors and different kinds of commercial and research applications.
These developments – significant price drops combined with improved capacity and efficiency, are in line with an analysis of the density and capability of hard drive storage media over time – are known as Kryder’s Law.
Now ten year’s old, Kryder’s Law is based on the work of electrical engineer Mark Kryder. Kryder spent several decades studying and observing the dramatic increases in magnetic disk areal storage density that have taken place over the last 60 years. During this time, period hard drives progressed from being able to store just several thousand individual bits of information in the early post WWII period – when dramatic developments in modern Computing were made – to the smaller, high-volume drives akin to those available today.
The concept of Kryder’s Law was conceived in 2005 when Kryder noticed that magnetic disk areal storage density was increasing in correlation to Moore’s Law, a computing theory that was originally created by Gordon Moore, Co-Founder of Intel.
Moore’s Law states that the number of transistors on a dense integrated circuit will double approximately every two years. Conceived in 1965 in Moore’s internal paper The Future of Integrated Electronics, the theory originally stated that the number of transistors would double every year, however, this was revised in 1975 and so far the prediction has proved accurate. Moore’s Law is now used by the semiconductor industry to guide long-term planning and to set R&D targets. Moore’s Law is seen as integral to the IT industry.
Although Kryder’s Law has since proved to have fallen a little short of the mark, it still reflects the technological change flash has contributed in terms of size and power of both consumer devices and the backend performance that sits behind programmes, websites and applications. Flash is thus helping to change user expectations of how fast and responsive their applications or websites work.
Flash and the data centre
The data centre is certainly one place flash storage technology is transforming the nature of data storage. The market for Solid-State-Array (SSA) set ups in data centres is expanding as enterprises invest more and more into the architecture, which will accelerate their workloads. SSAs enable this because they have the ability to provide increased performance in comparison to storage arrays that rely on spinning disk hard drives. They have much higher input/output (I/O) than their moving cousins. SSAs are often laid out as the top tier in an automated storage tiering approach and can decide where to move data based primarily on input/output activity, and thus does not prioritise those choices based on the application.
The migration to solid-state technology and software-defined storage approaches is inevitable because all-flash storage changes the way companies do business by dramatically speeding up application performance and improving server efficiency.
The next ten years
Kryder’s Law shows no sign of becoming irrelevant anytime soon. The price of flash is expected to continue to decline over time, and one of the most important factors is likely to be the development of new architectures, particularly the adoption of 3D NAND technology.
3D NAND, or “V NAND” enables memory cells to be stacked on top of each other vertically instead of spread horizontally across a grid. It allows flash memory to be expanded solving one of the drawbacks flash had in the early days, limited storage density.
While the future of flash storage, particularly for the next ten years, is likely to prove that Kryder’s Law is accurate, it doesn’t spell the death of HDDs. There are some workloads that will continue to run just fine on HDD arrays so don’t expect them to disappear from the data centre just yet.
Flash, however, is only going to become more welcome, especially in enterprise environments, where price has so far limited the advancement for industrial computing. While flash prices have traditionally been 5-10 times more expensive to do the same job: storing the same data as traditional HDD set ups, the declining prices of SSDs, relative to Kryder’s Law, is leading to a much more efficient data centre storage options, boasting up to ten times faster data retrieval while using a tenth of the power. As a result, flash will sooner or later become a mainstream technology in the data centre.
In 2005, Kryder said that since the introduction of the disk drive in 1956, the density of information it can record has swelled from just 2,000 bits to 100 billion bits, all crowded in the small space of a square inch, representing a 50-million-fold increase.
From this, Kryder theorised that because that 50 million increase took place over 41 years, it demonstrates the doubling the power and memory of computer semiconductors every 18 months, a rate of improvement faster than Moore’s Law.
Development of cost per Gigabyte of storage:
· Late 1950s: $10,000,000
· Late 1970’s: $500,000
· 1980: $200,000
· 1985: $71,000
· 1990: $9,000
· 1995: $686
· 2000: $7.14
· 2005: $0.58
· 2010: $0.07
· 2015: $0.03