The tremendous growth in flash storage we’ve seen over the last few years is likely to continue – with Gartner’s recent prediction that “by 2021, 50% of data centers will use SSAs (solid state arrays) for latency-sensitive workloads, up from less than 30% today.”1 As acquisition costs have fallen, all-flash has become accessible for more and more organizations to process huge quantities of input/ output operations, with many seeing a ROI in under six months.
But while businesses have been reaping the benefits of flash storage, they have been unable to maximize its full potential, as the traditional protocols that handle data flows between servers (or virtual machines) and storage arrays have throttled transfer speeds. These old school interfaces are serial technologies able to send or read data one bit at a time. Meanwhile, use cases are increasingly pushing these limits – with online transaction processing, data mining, real-time analytics and video editing and other applications that require parallel processing capabilities becoming commonplace.
The entire industry has been working on ways to address this challenge for some time. One key solution is the deployment of non-volatile memory express storage (NVMe). This has been developed from the ground up to deliver high bandwidth and low latency storage access for current and future non-volatile memory by replacing the traditional protocols that are restricting the potential of Flash. NVMe has also been designed with parallelism in mind since it supports far deeper queues (where storage commands wait to be executed) and more of them – up to 64,000, with up to 64,000 commands per queue.
Systems built to use the latest NVMe protocols from end to end – from server to storage – deliver the greatest performance benefits. While fully standards-based solutions are only just becoming commercially available, their adoption is likely to increase quite rapidly. Besides exceptional performance, these systems will have all of the capabilities and storage services that today’s demanding applications require – including high resilience and the ability to withstand component failure without restricting access to data or seriously degrading performance. They will also provide the data integrity, data encryption and system level security capabilities that many applications demand.
NVMe: The best is yet to come
Using NVMe systems brings benefits including the ability to more effectively utilize technology – from fully exploiting flash resources and taking full advantage of processor advances. System management is also streamlined thanks to features including a simplified command set and the ability to mix multiple workloads. It also delivers performance consistency and can support large data sets.
The industry’s storage and system vendors are already heavily committed to NVMe, and it is likely to become the storage standard for the next decade. As a result, it is already part of servers running workloads that demand the best performance. It is also starting to move into mainstream storage systems, and, as the ecosystem continues to expand, we’ll see more pure-NVMe systems.
Consequently, businesses should start planning to include NVMe in their future technology plans to ensure that their business-critical data continues to be processed as efficiently as possible. What is more, NVME was designed to be optimized without the need to rewrite applications – making it inherently future-proof. Its design also allows it to work with all device types, from mission-critical data center and cloud storage platforms to the storage in your tablet and smartphone. Moreover, its cost will start to fall as volumes increase.
To find out more about whether NVMe is right for your businesses, we’ve created a whitepaper in conjunction with the analyst firm Freeform Dynamics: “NVMe: a fresh start for storage”
Download it here.
1 Gartner, “2018 Strategic Roadmap for Storage”, 12 March 2018, ID G00350720