For decades, advances in storage technology have been measured primarily in terms of storage capacity and read/write speed. Over time, these evaluation parameters have been supplemented by technologies and methodologies that make HDDs and SSDs smarter, more flexible, and easier to manage. Every year, drive manufacturers traditionally hint that the big data market will change, and 2020 is no exception. IT leaders are looking for efficient ways to store and manage massive amounts of data, and as a result, they are once again promising to change the course of storage systems. In this article, we have collected the most advanced information storage technologies, as well as talk about the concepts of futuristic drives, which have yet to find their physical implementation.
Software Defined Storage Area Networks
When it comes to automation, flexibility and increased storage capacity, coupled with increased workforce efficiency, more and more enterprises are considering moving to so-called software-defined storage networks or SDS (Software-Defined Storage).
The key feature of SDS technology is the separation of hardware from software: that is, it implies . In addition, unlike conventional network-attached storage (NAS) or storage area network (SAN) systems, SDS is designed to run on any standard x86 system. Quite often, the goal of deploying an SDS is to improve operating costs (OpEx) while requiring less administrative effort.
The capacity of HDD-drives will grow to 32 TB
Traditional magnetic drives are not dead at all, but are just experiencing a technological renaissance. Modern HDDs can already offer users up to 16 TB of data storage. Over the next five years, this capacity will double. At the same time, hard disk drives will continue to be the most affordable random access storage and will retain their leadership in price per gigabyte of disk space for many more years.
Capacity expansion will be based on already known technologies:
- Helium drives (helium reduces aerodynamic drag and turbulence, allowing you to install more magnetic platters in the drive; at the same time, heat dissipation and power consumption do not increase);
- Thermal magnetic drives (or HAMR HDD, which are expected to appear in 2021 and are built on the principle of microwave data recording, when a section of the disk is heated by a laser and remagnetized);
- HDD based on tiled recording (or SMR drives, where data tracks are placed on top of each other, in a tile format; this provides a high density of information recording).
Helium drives are especially in demand in cloud data centers, and SMR HDDs are optimal for storing large archives and data libraries, accessing and updating data that is not required very often. They are also ideal for creating backups.
NVMe drives will become even faster
The first SSDs were connected to motherboards via a SATA or SAS interface, but these interfaces were developed more than 10 years ago for magnetic HDDs. The modern NVMe protocol is a much more powerful communication protocol designed for systems that provide high data processing speed. As a result, at the turn of 2019-2020, we see a serious drop in prices for NVMe SSDs, which become available to any class of users. In the corporate segment, NVMe solutions are especially valued by those enterprises that need real-time big data analytics.
Companies such as Kingston and Samsung have already shown what enterprise users can expect in 2020: we are all looking forward to the arrival of NVMe SSDs with PCIe 4.0 support, which will allow you to add even more data center speed when working with data. The declared performance of new products is 4,8 GB / s, and this is far from the limit. Next generations will be able to provide throughput at the level of 7 GB / s.
Together with the NVMe-oF (or NVMe over Fabrics) specification, organizations will be able to create high-performance storage networks with minimal latency, which will compete with direct-attached DAS (or Direct-attached storage) data centers. At the same time, using NVMe-oF, I / O operations are processed more efficiently, while the delay is comparable to DAS systems. Analysts predict that the deployment of NVMe-oF systems will rapidly accelerate in 2020.
QLC-memory will finally βshootβ?
NAND flash memory Quad Level Cell (QLC) will also show growing popularity in the market. QLC was introduced in 2019 and therefore had minimal market penetration. This will change in 2020, especially among companies that have adopted LightOS Global Flash Translation Layer (GFTL) technology to overcome QLC's inherent challenges.
According to analysts' forecasts, QLC-based SSD sales growth will increase by 10%, while TLC solutions will "capture" 85% of the market. Like it or not, QLC SSD is still far behind in performance compared to TLC SSD and will not become the basis for the data center in the next five years.
At the same time, the cost of NAND flash is expected to rise in 2020, so SSD controller vendor Phison, for example, is betting that higher prices will eventually push the consumer SSD market towards 4-bit flash. - QLC NAND memory. By the way, Intel plans to launch 144-layer QLC solutions (instead of 96-layer products). Well⦠it seems that further marginalization of the HDD awaits us.
SCM memory: speed close to DRAM
The widespread adoption of SCM (Storage Class Memory) memory has been predicted for several years, and 2020 may be the starting point in which these predictions will finally come true. While the Intel Optane, Toshiba XL-Flash and Samsung Z-SSD memory modules have already entered the enterprise market, their introduction has not caused an overwhelming reaction.
The Intel device combines the characteristics of fast but unstable DRAM with slower but persistent NAND storage. This combination aims to improve the ability of users to work with large data sets, providing both DRAM speed and NAND capacity. SCM is not just faster than NAND-based alternatives, it's dozens of times faster. The latency is microseconds, not milliseconds.
Market experts note that data centers planning to use SCM will be limited by the fact that this technology will only work on servers using Intel Cascade Lake processors. However, in their opinion, this will not be a stumbling block to stop the wave of updates to existing data centers in order to provide high speed information processing.
From foreseeable reality to the distant future
For most users, storing data doesn't feel like a capacitive Armageddon. But just think about it: The 3,7 billion people who currently use the Internet generate about 2,5 quintillion bytes of data every day. More and more data centers are needed to meet this need.
According to statistics, by 2025 the world is ready to process 160 Zetabytes of data per year (that's more bytes than there are stars in the observable universe). It is likely that further we will have to cover every square meter of planet Earth with data centers, otherwise corporations simply will not be able to adapt to such a high growth of information. Or⦠you will have to give up some data. However, there are several potentially interesting technologies that could solve the growing problem of information overflow.
The structure of DNA as the basis for future data warehouses
Not only IT corporations are looking for new ways of storing and processing information, but also many scientists. The global task is to ensure the preservation of information for thousands of years. Researchers at the ETH Zurich (ETH Zurich, Switzerland) believe the solution lies in the organic storage system that exists in every living cell: DNA. And most importantly, this system was βinventedβ long before the advent of the computer.
DNA strands are very complex, compact and incredibly dense as information carriers: according to scientists, 455 Exabytes of data can be recorded in a gram of DNA, where 1 Ebyte is equivalent to a billion gigabytes. The first experiments have already made it possible to record 83 Kb of information in DNA, after which the teacher of the Department of Chemistry and Biological Sciences, Robert Grass, suggested that in the new decade, the medical field needs to be more tightly integrated with the IT structure for joint development in the field of recording technologies and data storage.
According to scientists, organic data drives based on DNA chains could store information for up to a million years and accurately provide it on the first request. It is possible that in a few decades, most drives will compete for precisely this opportunity: the ability to reliably and capaciously store data for a long time.
The Swiss aren't the only ones working on DNA-based storage systems. This question has been raised since 1953, when Francis Crick discovered the DNA double helix. But at that moment, humanity simply did not have enough knowledge for such experiments. Traditional DNA storage thinking has focused on the synthesis of new DNA molecules; matching a sequence of bits to a sequence of four DNA base pairs and generating enough molecules to represent all the numbers that need to be stored. So, in the summer of 2019, engineers from the CATALOG company managed to write 16 GB of the English-language Wikipedia into DNA created from synthetic polymers. The problem is that this process is slow and expensive, which is a significant bottleneck when it comes to data storage.
Not DNA Aloneβ¦: Molecular Storage
Researchers from Brown University (USA) say that the DNA molecule is not the only option for molecular data storage for up to a million years. Low molecular weight metabolites can also act as organic storage. When information is written to a set of metabolites, the molecules begin to interact with each other and produce new electrically neutral particles that contain the data recorded in them.
By the way, the researchers did not stop there and expanded the set of organic molecules, which made it possible to increase the density of the recorded data. Reading such information is possible through chemical analysis. The only negative is that the implementation of such an organic drive is not yet possible in practice, outside of laboratory conditions. This is just an advance for the future.
5D optical memory: a storage revolution
Another experimental repository belongs to developers from the University of Southampton (University of Southampton, England). In an effort to create an innovative digital storage system that can last for millions of years, scientists have developed a process for writing data onto a tiny quartz disk that is based on femtosecond pulse recording. The storage system is designed for archiving and cold storage of large amounts of data and is described as five-dimensional storage.
Why five dimensional? The fact is that information is encoded in several layers, including the usual three dimensions. Two more dimensions are added to these dimensions - size and orientation by nanodots. The data capacity that can be written to such a mini-drive is up to 100 Petabytes, and the storage life is 13,8 billion years at temperatures up to 190Β°C. The maximum heating temperature that the disk can withstand is 982 Β° C. In short ... it is almost eternal!
Recently, the work of the University of Southampton has attracted the attention of Microsoft, whose cloud storage program Project Silica aims to reimagine current storage technologies. The "small-soft" predicts that by 2023, more than 100 Zettabytes of information will be stored in the clouds, so even large-scale storage systems will face difficulties.
For more information about Kingston Technology products, please visit the company's official website.
Source: habr.com