NVM Express Adds Computational Storage Feature
Mitch Lewis, Research Analyst at The Futurum Group, shares his insights on NVMe computational storage.
Cite this

The News: The NVM Express (NVMe) consortium recently announced a new addition to its specifications by adding a new computational storage feature. The new feature creates a vendor-neutral framework for using computational storage devices. More information can be found in the NVM Express press release here.
NVM Express Adds Computational Storage Feature
Analyst Take: NVMe, the consortium that oversees the NVMe specifications for SSDs, announced a new feature for computational storage. The feature adds new command sets and looks to set a standardized framework for connecting applications to computational storage devices.
Computational storage devices are, as the name strongly suggests, storage devices that contain built-in computational capabilities. While typically the term computational storage applies to devices with directly attached processors, in some cases, it may more generally apply to off-device accelerators that achieve similar outcomes. The overall idea behind computational storage is offloading specific computations from central processors to the storage devices themselves. This method can provide greater efficiency by reducing CPU workloads and by removing the overhead of data movement. Typically, computational storage devices have been used for applications such as encryption or data reduction. While the technology offers a compelling benefit in removing data movement bottlenecks, the overall adoption of the technology has been relatively slow.
The new NVMe computational storage feature looks to provide a vendor-neutral framework for connecting computational devices to applications. NVMe Computational Storage builds upon SNIA’s previously released Computational Storage Architecture and Programming Model with a specific focus on the NVMe specification. Included in the specification are two new command sets:
- The Computational Programs Command Set: Manages computational programs on the device and includes commands for loading, activating, and executing programs as well as creating and deleting memory ranges.
- The Subsystem Local Command Set: Supports access of memory in an NVMe subsystem via computational programs and NVMe transport. Commands include memory read, write, and copy commands.
NVMe’s addition of its computational storage feature puts a light back on a technology that seems to have struggled with adoption. While computational storage’s big selling point is reduction of data movement to increase performance, it faces competition from other approaches such as DPUs and other accelerator cards that solve similar problems. The new NVMe specification will help simplify and standardize utilization of computational storage devices, which may help boost adoption.
The new feature may also broaden the use cases of computational storage devices. While the typical use of these devices can certainly offer benefits and increase system performance, the typical program set has been fairly narrow, focusing on areas such as encryption, data reduction, or erasure coding. This may be in part by the computational power of the embedded devices, as well as a lack of defined and available programs. The new NVMe Computational Storage feature includes support to download programs. This ability may help broaden the applicability of such devices, especially as more computationally powerful devices continue to be developed.
The new NVMe Computational Storage feature is a positive development toward increased adoption of computational storage and ultimately removing data movement bottlenecks. One area in particular that may benefit heavily from broader use of computational storage in the future is the edge. The impact of this NVMe Computational Storage feature in boosting real adoption of the technology remains to be seen, but it will be an interesting area to keep an eye on.
Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.
Other Insights from The Futurum Group:
Seagate Announces Mozaic 3+ Drive Platform
Subscribe to The Weekly
The week's market signals, one email, no filler. Unsubscribe any time.
More from Mitch Lewis
Hammerspace Tier 0: Unlocking Greater Efficiency in GPU-Driven Computing
In our latest Research Brief, Hammerspace Tier 0: Unlocking Greater Efficiency in GPU-Driven Computing , The Futurum Group explores how organizations can overcome latency and storage inefficiencies by unlocking stranded NVMe capacity within GPU servers.
Compressing Storage Costs: How Modern Solutions can Lower TCO – Six Five On the Road
Tom Nagelmeyer and John Nicholson, Product Marketing Engineers at Broadcom, join Mitch Lewis and Camberley Bates to share their insights on how modern storage solutions are revolutionizing the way businesses manage and reduce the Total Cost of Ownership (TCO) for data storage.
Is VAST Data’s New InsightEngine the Solution for Enterprise RAG?
Mitch Lewis, Research Analyst, Camberley Bates, CTA, and Mitch Ashley, CTA, at The Futurum Group share their analysis on the VAST Data’s InsightEngine with NVIDIA announcements.