The world we live in is changing, and in recent years storage technologies are struggling to keep up. Just a couple of years ago, the IDC made the prediction that the average individual will have interactions with a connected device of some sort nearly 5,000 times per day by 2025.
This represents an increase of an order of magnitude over the nascent IoT networks of a decade ago, and yet in many instances the storage media that sit behind these networks are ancient by IT standards — NVMe-attached solid-state drives, or even “traditional” magnetic tape memory.
These hardware limitations are now giving rise to a wave of innovation. Some of this is focused at the managerial level, where memory management systems are allowing firms to deploy more sophisticated techniques for managing their data. Others are occuring at the level of fundamental technologies; advances in the ways in which we store data.
The emergent issue with our storage infrastructure can be put simply: it is no longer fast enough to keep up with increased computational performance, particularly in IoT and other edge-focused applications.
Up until now, engineers and systems architects have essentially made use of two types of storage: slow but permanent solid state drives, and DRAM. DRAM is fast enough to keep up with contemporary storage requirements, but unfortunately it is very limited in capacity in the average device. This has led to an unfortunate set of compromises, in which systems are either expensive, slow, or both.
Intel’s new Optane technology promises to address this issue by bridging the gap between DRAM and SSDs. Optane aims to accomplish this by stacking its memory cells, which are individually addressable, in a three-dimensional matrix. Intel is, rightfully, hesitant to share much more than this about the way that the technology works. However, early results are promising. Optane appears to be able to match DRAM for speed, while also allowing storage arrays to be switched off in the same way as persistent SSDs.
Optane is also well positioned to meet with success because it is highly backwards compatible. At the moment, Optane modules are available in capacities of up to 512GB, but with the possibility of going much larger in future. Because of this, we expect to see Optane being used in the top data management platforms and systems within a few years.
Also read: Top Data Management Platforms & Systems 2021
A more exotic form of future memory technology comes in the form of MRAM. MRAM stands for magneto-resistive RAM, and there is strong evidence that it can be used across a range of IoT devices. Since the IoT is going to be where the majority of extra storage demand is going to occur in the next few years, this could lead to rapid adoption.
The way in which MRAM works is highly technical, but the basics are easy to grasp. In “traditional” RAM, known sometimes as SRAM memory, data is stored on transistors inside RAM chips. Specifically, data is encoded in cells, each of which can hold six transistors. This works fine for some applications, but the high power requirements of SRAM mean that the memory is very inefficient.
SRAM stores data in a different way. It makes use of a pair of ferromagnetic plates, which are each divided by a dielectric insulator. Think of an SRAM chip as a kind of complex capacitor, and you’re on the right lines. By polarizing the ferromagnetic plates, data can be stored in MRAM chips in a way that is persistent, like SSDs, but requires a fraction of the energy of SRAM.
If MRAM seems set to change the way in which IoT data is stored, ReRAM might do the same for the data center. However, the advantages of ReRAM over traditional SSD storage is not lower power consumption, but instead increased speed.
The way that ReRAM works bears a lot in common with standard NPN transistors. However, ReRAM cells can be made much smaller than standard transistors, which are now getting so tiny that quantum effects are beginning to undermine their performance. ReRAM allows for multi-level chips that have greatly increased data density, and therefore greatly increased speed.
Initially, the cost of ReRAM is likely to limit adoption. However, some companies are already considering how to integrate ReRAM into their systems-on-a-chip, making use of the increased speed of ReRAM but not relying on it for mass data storage.
Whichever of these technologies gains prominence over the next few years, it’s already apparent that there is a huge market for innovative storage solutions. Coughlin Associates predicts that 3D XPoint memory, the technology at the heart of Optane, will drive revenues to over $16 billion by 2028.
Alongside these shifts, we’re also seeing an increasing number of companies who are aware that it’s possible to gain a competitive edge using data management, and this will drive adoption even further. In short, with the three technologies we’ve mentioned above, we might finally be reaching the stage where our storage technologies can keep up with our computing power.