Skip to content Skip to footer

The NVMe flash revolution

Image: Adobe Stock / pinkeyes

In data centres the world over, IT teams are working hard to balance both storage performance and sustainability. This brings with it some important technology choices, with NVMe flash – a high-performance, non-volatile storage technology – becoming an increasingly popular alternative to traditional disk-based solutions in meeting today’s demanding requirements.

As well as its adoption across high-performance use cases, NVMe flash is also being increasingly utilised as a cost-effective solution for cold data storage. To make the most of it, however, organisations also need a storage platform running at the software layer to take full advantage of its capabilities.

Indeed, legacy HDD technologies are already being phased out, driven in part by the cost per gigabyte price of flash storage, which is expected to be lower than higher-end HDDs during this year. This is the latest milestone in a trend which has seen the cost of flash decrease over the past 15 years or so, with flash-capacity shipments growing by nearly 40% last year alone as a result.

What’s more, last year, the volume of shipped HDD capacity decreased for the first time on record, while at the same time, the ratio and capacity of flash and tape both grew. Looking ahead, this arguably signals a direction of travel whereby HDD will be squeezed out of the market by flash from the top and by tape from the bottom.

These market pressures are likely to accelerate the adoption of NVMe-based solutions even further, thanks to its potential to deliver the highest performance levels available across current storage technologies. Given the exponential growth in the collection, storage, and use of unstructured data, legacy storage systems increasingly don’t have the capability to support workloads such as those seen across Artificial Intelligence (AI) and Machine Learning (ML) applications.

Overcoming the inefficiencies of SATA

There are also other important issues driving the growth of NVMe. In most contemporary data centres, many SSDs still use the inefficient Serial Advanced Technology Attachment (SATA) standard for connecting and transferring data to and from computer systems. As SSDs have become faster and IT teams try to squeeze every ounce of performance out of their storage systems, many are turning to NVMe as it was specifically designed to work with solid-state memory.

In particular, its design allows host hardware and software components to calculate processes simultaneously. This not only reduces I/O overhead but does so while improving performance and reducing latency. This makes it possible to process millions of concurrent operations – a huge performance improvement over legacy architectures using SAS or SATA.

This approach has already delivered significant performance benefits for applications that require faster read-/write-speeds for large files or lower rendering times than traditional technologies can deliver, such as higher resolution 4K and 8K video content. As performance requirements keep on rising across the board, everything from AI/ML, business intelligence, data analytics, and DevOps use cases (among many others) are driving demand for NVMe all-flash architectures.

And almost everywhere else, enterprises are looking to the capabilities offered by NVMe flash to help deliver on their Environmental, Social, and Governance (ESG) targets. By reducing the amount of hardware required, for instance, it becomes possible to reduce this element of infrastructure footprint by as much as 80% while also consuming significantly less power.

Unlocking NVMe’s potential

Despite the clear and proven advantages offered by NVMe flash legacy solutions, taking full advantage of their capabilities relies on integrating the technology with a modern storage platform at the software layer. For example, current NVMe-based storage solutions often require custom hardware and, as a result, will not run in the cloud, meaning many users can’t fully benefit from falling costs.

Instead, implementing a software-defined storage solution (SDS) developed specifically for NVMe can solve this problem by separating hardware and software layers. As such, it opens up a pathway for enterprises to benefit from the best possible solutions via a modern, cloud-native distributed file and object storage platform, purpose-built for NVMe.

By unlocking the true potential of flash, these architectures also eliminate the performance bottlenecks associated with legacy technologies. In doing so, organisations trying to cope with rising volumes of unstructured data can deliver the performance they need from their storage systems.

In an era of digital transformation and the rapid adoption of resource-hungry AI and ML systems, IT teams that focus on the benefits a modern, NVMe-compatible storage solution has to offer will be ideally placed to succeed.

Picture of Skip Levens
Skip Levens
Director, Media & Entertainment at Quantum

You may also like

Stay In The Know

Get the Data Centre Review Newsletter direct to your inbox.