image_pdfimage_print

Artificial intelligence (AI) has continued to grow and spread across industries to become one of the fastest-growing and largest investment areas. In the 2021 Forrester report, “Solving the Challenge of Enterprise AI Infrastructure,” 88% of IT decision-makers surveyed reported that in the next two years artificial intelligence and machine learning (AI/ML) infrastructure is likely to be their largest investment area or a major one. 

AI can quickly analyze massive amounts of data, revealing correlations and patterns that would take humans months to derive. This is not just limited to large scientific institutions and academic research projects. There are so many practical applications for this technology that it’s no surprise AI is now supporting mainstream use cases in industries from healthcare and life sciences, to semiconductor and chip manufacturing, to automotive, financial services, and beyond.

Challenges of AI Implementation

Organizations that want to begin using AI to fuel faster insight must be ready to adapt quickly to how new technology works as well as maintain it. The IT infrastructure historically used to support legacy data science efforts can’t provide data access that is fast enough to unleash the advantage of GPU-powered AI-based data science.

Organizations across virtually every industry are eager to learn how AI can empower their business and drive innovation. However, successfully implementing and using AI is not as easy as installing a program or buying equipment, and organizations need help to make this transition as smooth as possible.

Introducing AIRI//S

Pure Storage® is taking AIRI, the AI-ready infrastructure, to the next level with AIRI//S to help organizations simplify implementation and capitalize on the capabilities of AI. Built upon Pure’s years of proven leadership in storage for AI deployments and architected in collaboration with NVIDIA, AIRI//S consists of the NVIDIA DGX A100 system, the latest DGX™ system available in the market, the NVIDIA Spectrum Ethernet storage fabric, and the new FlashBlade//S, the cutting-edge and modular scale-out storage platform for unstructured data from Pure Storage. The result is an incredibly easy-to-use and easy-to-set-up, out-of-the-box AI solution that delivers prodigious performance from day one and flattens the barriers to entry for AI.

AIRI//S provides a simple and highly scalable architecture that is easy to set up. AIRI//S is built with the NVIDIA DGX software stack, the Pure Storage Purity//FB operating system, Pure1® cloud management, and the RapidFile Toolkit. This architecture is optimized across compute, networking, and storage to help you consolidate multiple hardware silos into one shared AI infrastructure capable of supporting multiple AI data science initiatives and pipelines. AIRI//S offers:

  • Ease of use: AIRI//S can be set up, deployed, and optimized quickly as an end-to-end AI pipeline solution. It ensures that teams are not overwhelmed by the infrastructure requirements for AI and data teams to remain productive, enabling them to spend more time running experiments and building models rather than worrying about their infrastructure.
  • Performance at any scale: AIRI//S is optimized out-of-the-box to deliver multi-dimensional performance for all types of GPU-powered workflows. You can seamlessly scale its storage and compute capabilities which allows organizations to start their AI journey at any scale and grow at their own pace. 
  • An agile AI platform: AIRI//S is built with future innovations from Pure and NVIDIA in mind. FlashBlade//S is designed as always-modern storage infrastructure that accelerates future innovation, making it simpler to bring denser, more power-efficient, and more performant capabilities to market faster and non-disruptively to address the evolving demands of customers’ AI requirements. FlashBlade//S is designed to support the latest NVIDIA DGX systems and related AI ecosystem technologies.  
  • Sustainable and efficient AI infrastructure: FlashBlade//S enables a more powerful AI infrastructure in a smaller footprint by delivering higher density while reducing power consumption and providing unmatched data center efficiency.

The New FlashBlade//S

FlashBlade//S delivers new levels of performance, density, and power efficiency to AIRI//S, making it the ideal storage platform to pair with the industry-leading performance of NVIDIA DGX A100 systems. The modularity of FlashBlade//S architecture and the future-proof nature of each component makes it easy to deploy the latest innovation in the industry and keep up with evolving customer requirements. When it’s time to grow, FlashBlade//S is upgradable on the fly, with no interruption to AI operations. FlashBlade//S provides greater capacity (up to 1.92PB/chassis) and performance (greater than 2x increase over the previous-generation FlashBlade®), all while also consuming 48% less power (compared to alternative scale-out NAS platforms).

NVIDIA DGX A100

NVIDIA DGX A100 systems include eight NVIDIA A100 Tensor Core GPUs and up to ten NVIDIA ConnectX-6 network adapters. NVIDIA Spectrum Ethernet Storage Fabric (ESF) delivers the performance and feature set needed for high-performance storage networking. Additionally, with NVIDIA MIG technology, users can partition each A100 GPU into up to seven instances for efficiency across large AI training and smaller inference workloads.

Pure Storage: Enables Customer AI Success with NVIDIA

Pure Storage has worked with NVIDIA to help customers realize their AI initiatives across a variety of industries, including the federal and public sector, healthcare, manufacturing, financial services, and telecommunications. AIRI//S speeds time to insight by removing design complexity while providing performance scalability in an agile, efficient, and future-ready AI platform.

Here are examples of the success Pure and NVIDIA have enabled with customers across multiple industries:

  • Advancing the national security mission of the United States and other nation’s intelligence agencies (private references).
  • Accelerated genetic research and medical imaging 7X at the Center for AI in Medicine at Chang Gung Memorial Hospital
  • Accelerated medical image processing, increased customer satisfaction and market competitiveness at Olympus, a leading MedTech provider.
  • Accelerated training models to advance the capabilities and safety of self-driving vehicles for Zenuity.
  • NavInfo relies on AIRI for advanced AI analysis of large data sets to produce real-time maps and navigation models used by car manufacturers around the world.
  • Delivered significant expansion in the volume data science research projects for the Gwangju Institute of Science and Technology (GIST)
  • Accelerating EDA to support the pace of chip development where for each nanometer a die is reduced in size, the volume of data required to be processed in the development almost doubles from the prior size (private references).

How Can We Help?

AIRI//S from Pure Storage powered by NVIDIA accelerates and simplifies the process of deploying and running complex AI-ready infrastructure so our customers’ data teams can focus on their core mission and discover critical insights and outcomes faster. Learn more about the AIRI//S solution.