By David Cohen, System Architect and Senior Principle Engineer at Intel
With the arrival of new non-volatile memory (NVM) technologies, we are suddenly in the midst of the biggest data center transformation in the past 30 years. Data centers are now poised to move data at unprecedented speeds.
This isn’t hyperbole. This is the way it will be with the implementation of solutions built around the new technologies like NVM Express* (NVMe) Over Fabrics, NVMe in PCI Express, and 3D XPointTM. These technologies will bring down the cost of non-volatile memory and replace hard disk drive (HDD) with solid-state storage—while taking storage performance to unprecedented levels.
A case in point: The new 3D XPoint (pronounced “3D cross-point”) technology from Intel and Micron enables NVM speeds that are up to 1,000 times faster than NAND, today’s most popular non-volatile memory. With its unique material compounds and cross-point architecture, 3D XPoint technology is 10 times denser than conventional memory. We’re talking about a category of NVM that has the potential to revolutionize any device, application, or service that can benefit from fast access to large sets of data.
Let’s take a step back and look at the bigger picture. In enterprise data centers, spinning disk drives (HDDs), which continue to carry a lot of the data storage load, have always been really slow (in relative terms) while everything else has been really fast. This amounts to a bottleneck in the application performance pipeline. At some level, it doesn’t matter how fast today’s processors and network switches are when overall performance is tied to the speeds of yesterday’s data storage devices.
We are now in the process of rewriting this tired equation. With the arrival of solutions based on the new non-volatile storage technologies, storage will be really fast and, in comparison, everything else will be slower. For the software developer, this new reality of lightning-fast storage creates an imperative to optimize other parts of performance pathway to remove overhead that causes latency.
In essence, the goal is to move operations that are not critical to performance out of the performance path—such as bookkeeping functions related to the management of transactions and data replication operations that could take place elsewhere. The idea is to tease latency out of the system and allow all things to happen in parallel. The ultimate goal is a balanced system that capitalizes on the full potential of the latest server, storage, and networking components.
At Intel, we are committed to making our end customers successful in the transformation to next-generation silicon-based storage. To that end, we are working actively with our ecosystem partners and end-user customers to help ensure that software is addressed in the right way—so that operating systems and applications can gain the greatest benefits from faster storage. At the same time, we are working closely with industry organizations to make sure there are standards in place that allow software developers and OEMs to capitalize on new storage technologies in a uniform way.
Here’s the bottom line: The future is upon us. It’s ours to make of it what we will. To maximize the potential of storage solutions, we must first embrace new technologies like NVM Express and Next Generation NVM, and then work actively to optimize the associated software for the new capabilities of NVM.
Our success in these efforts will throw open the doors to the next generation of data centers.
For a closer look at the new non-volatile memory technologies, and the future of storage itself, visit intel.com/storage.
1 Performance difference based on comparison between 3D XPoint technology and other industry NAND.
2 Density difference based on comparison between 3D XPoint technology and other industry DRAM.