Pure Storage announces new FlashArray//X series AFA and new AIRI AI ready systems Silverton Consulting, Inc. StorInt™ Briefing
At Pure//Accelerate 2018 this past month, Pure Storage announced an expanded FlashArray//X storage family and a new FlashBlade-based AIRI Mini appliance that integrates NVIDIA compute with Arista or Cisco networking to speed AI workloads and workflows.
New FlashArray//X series systems
Pure has revamped their FlashArray family from the SSD-based M series to offer a broad range of all-NVMe and NVMe-oF ready FlashAray//X products. Last year Pure Storage introduced the FlashArray//X70 with NVMe based DirectFlash internal storage media. This year they have refreshed the //X70 and added new FlashArray//X50 and //X90 systems as well as lower end SATA SSD, NVMe-capable FlashArray//X10 and //X20 storage.
All DirectFlash //X systems are NVMe-oF ready and designed to support new SCM, ultra-fast SSDs when available. All //X systems with DirectFlash also offer access to an optional DirectFlash Shelf storage connected via NVMe-oF. Moreover, with Evergreen storage all current and previous generation FlashArrays can be non-disruptively upgraded, in place, up to the highest performing FlashArray//X90 storage.
Purity FA5 storage software now supports up to 2X more write performance. FA5 also adds a new background compression engine that can reduce data footprint by up to 20% over previous FlashArray systems. Pure states that the new FA5 FlashArray//X storage can achieve an average of 5:1 data reduction. At 5:1, the new //X systems support 55TB, 275TB, 650TB, 1.3PB and 3PB of effective capacity for FlashArray//X10, //X20, //X50, //X70 and //X90, respectively.
The NVMe DirectFlash module is a Pure Storage designed, proprietary flash device which provides Purity FA5 with a global flash pool, with all raw NAND flash visible to FlashArray//X software. DirectFlash modules help FlashArray//X supply more deterministic, low latency IO and dense flash media, including up to 18.3TB of raw storage per module. With DirectFlash, Pure offers up to 43% more usable flash storage from the same quantity of NAND, giving Pure a pricing advantage over vendors that use off-the-shelf SSD. Moreover, FlashArray//X with NVMe DirectFlash provides up to 4X more IO performance density than SAS SSD storage, driving further cost advantage.
FlashArray//X series controllers each come with 4 25GbE ports onboard which can be expanded to include up to 6 25/40/50 GbE ports or up to 10 16GFC or 32GFC ports per controller or double those, at the system level.
Pure states that the new //X series storage systems with NVMe DirectFlash is up to 2X faster than previous generation FlashArray systems and has as low as 250 msec latencies, with always on, full data services and QoS.
The new FlashArray//X systems are priced to match previous generation FlashArray//M storage. Last year’s FlashArray//X70 25% price premium is now gone.
New AIRI Al ready infrastructure
Pure also announced AIRI Mini, which builds on the AIRI converged solution launched earlier this year. AIRI Mini combines two NVIDIA DGX-1™ systems (compared to 4 for the larger AIRI system), Pure FlashBlade storage and Arista or Cisco networking. The AIRI Mini system extends Pure’s converged infrastructure for AI to entry-level use cases.
NVIDIA DGX-1 system uses eight Tesla V100 GPUs for up to 4 PFLOPs of DL performance and supports NVIDIA optimized deep learning frameworks such as Caffe/Caffe2, Microsoft Cognitive Toolkit, Apache MXNet, Google TensorFlow, TheANO, and PyTorch. It also includes GPU optimized Docker support for containerization and scale out GPU compute.
Recent, Resnet-50 and ResNet-152 industry standard, AI benchmarks results have proven AIRI systems are faster than other leading-edge storage systems-DGX combinations
The Pure//Accelerate show this year was held at the Bill Graham Civic Auditorium in San Francisco. Pure said they had ~5K participants at the show with another up to 10K people viewing the live stream.
Pure is leading the pack with the AIRI Mini appliance and the AIRI architecture using their FlashBlade storage, which is highly parallel, low latency and seemingly tailor made for AI DL IO. Moreover, the new FlashArray//X systems using internal NVMe DirectFlash storage and NVMe-oF attached DirectFlash storage shelf, supplies a state of the art, NVMe storage for the upcoming market transition from All-Flash Arrays to Shared Accelerated Storage.
[This storage announcement dispatch was originally sent out to our newsletter subscribers in June of 2018. If you would like to receive this information via email please consider signing up for our free monthly newsletter (see subscription request, above right) and we will send our current issue along with download instructions for this and other reports. Dispatches are posted to our website at least a quarter or more after they are sent to our subscribers. Also we have recently updated (February 2019) our SAN Storage Buying Guide so if you are interested in purchasing block or file primary storage, please checkout any of our Buying Guides available for sale on our website. ]
Silverton Consulting, Inc., is a U.S.-based Storage, Strategy & Systems consulting firm offering products and services to the data storage community