Whenever there's a discussion of enterprise NVMe, at some point adoption rate becomes the focal topic. Sure, the adoption rate has been protracted, but the pace makes sense when you take a step back and consider all of the variables at play.

Even when you include the unmistakable fact that the global market is still adjusting from changes brought on by the pandemic, there is no single reason as to why NVMe adoption in enterprise environments has been drawn out.

NVMe is a relatively new storage protocol. It effectively replaces SAS and SATA as the primary standards for storage interfacing and also replaces non-standardised direct interfacing of SSDs via the PCIe bus. NVMe was specifically designed for the unique characteristics of SSDs and storage-class memory. Being originally developed to command spinning platter hard drives, SAS and SATA are simply too bloated and inefficient for SSDs.

Looking back at the history of SCSI, early adoption in the 1980s was modest. SCSI first took hold in workstations mainly with limited distribution in specialised server architectures. It took almost 20 years for SAS to evolve from the parallel SCSI-1 and Fast SCSI interface types.

Using SCSI as a model, it's far easier to understand why the deployment path of NVMe looks the way it does. It's been ten years since the v1.0 specification was released, it's only seven years since hardware started to appear in earnest and, just like SCSI, we're seeing it first in workstations and specialised server architectures.

So, why is NVMe adoption in enterprise architectures taking so long? According to a recent survey by Vanson Bourne, organisations adopting NVMe storage currently face two main challenges: scalability and cost. In addition, respondents indicated that they still need NVMe to offer greater flexibility and manageability.

Low-overhead NVMe technology looks highly promising for future wide-scale deployments, yet IT teams today require proven reliable and effective technologies. NVMe SSDs could see throughput rates of up to 32GB/s and some do reach 10 million IOPS. However, current limitations in hardware and application infrastructure make it difficult to reach this level of performance cost-effectively.

NVMe SSD prices can exceed ten times more per GB than traditional HDDs. As a result, the investment required to purchase NVMe All-Flash Arrays remains extremely high. What's more, management tools such as the NVMe Management Interface (NVMe-MI) specification are still in development and not yet widely deployed.

The good news is that solutions are on the way. These include PCIe 4 hardware with greater NVMe support, solutions for scaling NVMe and expanded and refined management applications. While it will take several more years for enterprise-grade NVMe SSD storage prices to come down, prices are slowly becoming more competitive.

Most IT decision-makers are unfamiliar with the total cost of ownership calculations of deploying NVMe SSDs versus SAS SSDs. This has been a limiting factor in determining how money is spent on hardware especially when price takes priority. Eventually, there will be a convergence of price, utility and performance not unlike what happened with SCSI and SAS.

All of this excludes the interesting developments in bringing together Fibre Channel and NVMe where the former is proving to be very useful in supporting networked NVMe storage. The merging of Fibre Channel with NVMe storage could go a very long way in expediting the appearance of a positive TCO simply because existing SANs can be left in place.

While NVMe adoption in enterprise environments has been taken its time, the pace does follow historical precedent. NVMe takes the lead in performance yet SAS currently holds the advantage in other key factors: cost, scalability, manageability, and support. Eventually, though, a balance will be reached where we will take the inclusion of NVMe for granted just as much as we do any other technology today.

More info: www.atto.com