There’s no argument that flash storage is in high demand for today’s hyperscale and cloud storage data centers. In fact, flash-based SSDs have a role in most datacenters and there is a strong desire for many hyperscalers to go all-flash as quickly as possible. In order to address as much of the market as possible, SSDs have evolved to be fairly generic and have very little differentiation. This is okay for the general-purpose workloads of some enterprise applications but this one-size-fits-all approach is not a good fit for the growing hyperscale and cloud datacenter segments. There is a better way to design storage that will accelerate the transition to the all-flash datacenter.
Over the past decade or so, everything in the cloud/hyperscale datacenter has been redesigned to fit these new architectures EXCEPT the storage components (HDDs and SSDs). Storage devices have the same architecture and delivery model they have always had. On the other hand, software-defined architectures have improved storage arrays, networking, security and most other aspects of the datacenter. It is now time for the storage components themselves to benefit from the flexibility and tunability provided by being software-defined.
In one of my previous posts, I explained the benefits of using programmable controllers to speed innovation, improve agility, and break down many of the technical and cost challenges that are impeding advancements in cloud and hyperscale storage. Obviously, programmable hardware is only a part of the equation and must be married to powerful software in order to deliver these benefits.
Specifically, Burlywood’s FlashOS™ software delivers powerful features that are unmatched by standard SSDs. As an example, FlashOS provides true workload matching where all aspects of the architecture are tuned to provide the highest performance as well as the most efficient use of the flash. This goes well beyond simple quality of service to include variations at the flash translation layer (FTL) that are customized to how the application interacts with the flash. Furthermore, if the application is modified over time, re-tuning to match those changes is as simple as a firmware download. Being software-defined means never being forced to compromise performance because you are stuck on ASIC-based storage controllers.
A software defined SSD architecture greatly decreases the time to market for new features and the time required to support new non-volatile memory technology. This is more important than ever in light of the current trend to accelerate the transition from one flash node to the next and it enables multi-sourcing at the Flash component level. Taking advantage of the latest flash technology before their competition gives FlashOS customers a competitive edge.
The concept of a software-defined datacenter is not new, but extending those advantages to the storage component level is a radical extension of that trend and, in our opinion, a powerful new tool in the transition to the all-flash datacenter.
Look us up on social media and reach out if you would like to know more about how the first software-defined SSD architecture based on Burlywood FlashOS is redefining how flash storage works.Tags: datacenter, flash, FPGA, NVMe, SSD