Hyperparallel Flash Arrays (HFAs) are the driving force behind the third wave of storage. HFAs leverage a fundamentally different architecture than the dual controller design of legacy arrays. HFAs, with multiple controllers, are storage arrays that can linearly and independently scale both capacity and performance to redefine what storage arrays are capable of doing in today’s modern data centers.
Legacy AFAs were designed for availability, and use the same dual controller architecture that was intended to meet the performance needs of an array of spinning disks. Based on the serial design of disks connected to a single active controller, this type of configuration can never fully take advantage of the parallel access that SSDs offer.
Just as NVMe was needed to replace SAS and SATA for SSDs, a hyperparallel architecture is required to replace the legacy design of AFAs.
This new architecture has allowed HFAs to unlock the parallel access capability of flash media. As a result, HFAs deliver dramatically improved performance, far beyond what legacy designs can offer. Where there were once two controllers, at best running half speed in an active-active configuration, there can now be 10 or 20 controllers, each operating at maximum capacity and accessing storage simultaneously. As requirements change, controllers and capacity can be added independently of each other, providing linear scalability.
Designed for scale
In the legacy array model, when an organization needed to scale it was usually done through the addition of an entire new array, which then used software to cluster the systems together. This meant that a complete new storage system was required for scale, whether all the resources of that array were needed or not. Independently scaling capacity and performance was rarely possible. This approach also assumed that the array could support being clustered with another system, a requirement that’s not a given.
HFAs, with the ability to use multiple controllers simultaneously, can independently scale capacity or performance on an as needed basis. If more performance is required, then additional controllers can be added. To grow capacity, either new drives can be added or larger ones can be used in place of smaller drives. If a new array is needed, it can simply be added to the storage environment without the high complexity of the past.
By taking advantage of their multi-controller architecture and the parallelism of flash, HFAs can dramatically outperform legacy dual controller designs. Consider that typical performance for a dual controller AFA is roughly 8 GB/s bandwidth, 2M IOPS, and 500μs latency. In contrast an HFA (such as Pavilion Data Systems) can deliver 90 GB/s bandwidth, 20M IOPS, and 40μs. This is the benefit of a parallel architecture, designed to unleash the power of SSDs. This is the power of a hyperparallel flash array. This is why the industry is now embracing the Third Wave of Storage.
Learn more about this subject by visiting NVMeAcademy.com