IBM SpectrumAI and NVIDIA DGX to Ease Deep Learning

IBM SpectrumAI and NVIDIA DGX to Ease Deep Learning

Modern companies are mostly driven by data and accompanied data insights. In order to derive meaningful insights, companies are aggressively investing in data scientist and powerful machines to develop AI/ML/DL, models. Thus to implement AI initiatives it is imperative to remove any hurdles and derive efficiency in AI data pipelines and maximize data science productivity.

Now IBM and NVIDIA, this week, partnered to offer converged infrastructure to deliver integrated compute, storage and networking to support AI lifecycle from data preparation to training to inference. A High-performance, multi-protocol shared storage is needed for the latest AI and data tools, like TensorFlow, PyTorch, and Spark.

With the release of IBM SpectrumAI with NVIDIA DGX, IBM has expanded on its IBM Systems AI Infrastructure Reference Architecture, improved AI development productivity, and streamlined AI data pipeline.

IBM SpectrumAI with NVIDIA DGX incorporates software-defined storage scale-out file system, Spectrum Scale on flash with NVIDIA DGX-1. Below are the key elements of this solution,

  • IBM Spectrum Scale v5: Software-defined to streamline data movement through the AI data pipeline
  • NVIDIA DGX-1 servers: a purposely-built for AI and machine learning
  • NVIDIA DGX software stack: Optimized for maximum GPU training performance
  • Proven data performance: Over 100GB/s throughput, supporting up to nine DGX-1 servers in a single rack

IBM Spectrum Scale can be configured from a single IBM Elastic Storage Server (ESS). It can be extended from a few NVIDIA DGX-1 servers to a rack of 9 servers with 72 Tesla V100 Tensor Core GPUs to multi-rack configurations.


The announcement noted, "Unlike traditional storage arrays, the highly parallel IBM Spectrum Scale scales practically linearly with random read data throughput requirements to feed multiple GPUs.The result is a solution that delivers AI workload performance from shared storage comparable to the that of local RAM disk. In a single reference architecture rack, IBM Spectrum Scale on three IBM NVMe arrays demonstrated 120GB/s of data throughput to support multiple user and multiple models simultaneously."

Senior Director of DGX Systems at NVIDIA, Charlie Boyle, said, “Our customers are looking to scale up their data science infrastructure to help solve the challenges of AI data deployments. The power of the GPU-accelerated NVIDIA DGX-1, combined with IBM Spectrum Scale’s storage technology, gives data scientists a turnkey solution that brings together industry-leading compute and high-performance storage with proven results.”

VP Offering Management, IBM Software-Defined Storage, Sam Werner, said,
“The same storage innovations that are driving the fastest and smartest supercomputers in the world are now available to any data science team building high-performance AI model training infrastructure. Integrated by our joint channel partners, IBM SpectrumAI with NVIDIA DGX can get our clients started quickly and help to grow their AI, machine learning and deep learning projects.”

The offering is available at IBM AI infrastructure portal. IBM and NVIDIA will be also conducting a webinar on 29th January 2019 and details can be accessed here.

Note: We at TechSutram take our ethics very seriously. More information about it can be found here.
Mandar Pise Opinions expressed by techsutram contributors are their own. More details

Mandar is a seasoned software professional for more than a decade. He is Cloud, AI, IoT, Blockchain and Fintech enthusiast. He writes to benefit others from his experiences. His overall goal is to help people learn about the Cloud, AI, IoT, Blockchain and Fintech and the effects they will have economically and socially in the future.

No comments:

Post a Comment

    Your valuable comments are welcome. (Moderated)