NVMe 101: What Is It, and Why Should You Care?

Rob Commins<br/>TegileRob Commins
Tegile

Rob Commins is VP of Marketing for Tegile.

Eight years after its introduction into the enterprise market, flash storage is finally hitting its stride. A recent survey from ActualTech Media shows that hybrid storage systems (a mix of flash storage and disk) are now found in 55 percent of data center environments, and almost one-third of data centers now have at least one all-flash array. This means that more than two-thirds of data centers now use flash technology. As storage prices continue to drop, adoption will continue to grow year-over-year.

However, despite the impressive speed and performance that flash storage brings to the data center, the reality is that businesses have hardly scratched the surface of its capabilities due to architectural limitations.

A new technology called Non-Volatile Memory Express (NVMe) is about to change that.

What is NVMe?

NVMe is a host controller interface and storage protocol designed specifically for solid state drives (flash) and any other persistent memory technology developed in the future. Currently, the two most common controller interfaces found in the data center – SAS and SATA – both act as bottlenecks to flash storage performance. This is because they were designed only with the performance characteristics of spinning disk in mind.

NVMe is designed from the ground up to take advantage of the unique capabilities of solid state drives, using Peripheral Component Interconnect Express (PCIe) as its serial expansion bus.

The development of NVMe began in 2009 as an effort to create a new industry standard for storage devices to operate with host computers. Seven years of hard work has resulted in numerous benefits compared to older common controller interfaces. Here’s what NVMe brings to the table:

Superior Performance

Most people don’t realize that flash storage is a parallel storage medium. Unlike SAS or SATA, which treat flash in a serial manner, NVMe is designed to take advantage of the awesome power of parallel processing. This means it can handle 64,000 queues of data, and each queue can process 64,000 commands – at the same time. To put this in perspective, SAS and SATA can only hold a single queue, with just 32 and 256 commands, respectively.

NVMe is also incredibly efficient. It’s I/O request path is much shorter than its predecessors. In fact, the amount of CPU commands is reduced by half, and it supports interrupt steering – meaning data transfers can intelligently overcome even the largest interrupt signals. This translates into lower latency on both the software and hardware level, and is less taxing on your power consumption.

An Architecture Built for Scaling

Every data center administrator is familiar with the perils of buying new storage: Instead of focusing simply on the best array for you, you have to address a labyrinth of questions. For example, will your array be compatible with your specific controller interface? Will it function with your operating system? Should you be thinking about a gradual pivot from SATA to SAS? Using NVMe means you no longer have to think about these questions. It is designed specifically for PCIe, and almost all vendors have agreed to adhere to NVMe protocols.

Make no mistake, the high IOPS and low latency of NVMe will deliver significant competitive advantages to any business, having a positive impact on everything from high performance computing, to virtualization and the private cloud. The price range is high at the moment, but just like flash (or any other innovative technology), it will commoditize and drop.

Case in point: G2M research predicts that NVMe will be a $57 billion market by 2020, with a 95% compound annual growth rate (CAGR). The way I see it, we’ve crossed the chasm, with over half of data centers deploying flash in some capacity. With NVMe finally here, the concept of an all-flash data center – once considered a pipe dream – is suddenly in view.

Opinions expressed in the article above do not necessarily reflect the opinions of Data Center Knowledge and Informa.

Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.

Add Your Comments

  • (will not be published)

One Comment

  1. Hi Rob, long time! What I usually ask my partners when explaining NVMe technology and the wonders of low latency is this: since NVMe bypasses the RAID controller in a server implementation of NVMe, how is data protection accomplished? The responses, besides crickets, is alarming. Of course, since you are from a storage company, we can assume data protection is in the array. But host based NVMe, where the perfromance is phenomenal? -- Well, I'm not going to give away my answer.... Maybe I should write an article....