Nvidia GTC: IBM with AI Storage and AI-Direct Data Accelerator

By David Wohlford, marketing manager, WW cloud storage portfolio, IBM Corp.

Data matters and as AI spending is expected to grow to $301 billion by 20261 according to IDC research and deploying infrastructure to support new AI workflows is the top business initiative.

When planning an AI infrastructure strategy, it’s easy to focus on compute resources and the applications without a strategy for the data that fuels the results and the benefits of AI. Careful thought and planning must be used when deploying an AI data strategy, but we find there are several challenges that provide roadblocks to success. 

With companies like Nvidia Corp. and their accelerated performance using GPUs there is a growing need to increase the performance of workloads and use more data from multiple sources that needs to be accessed. The problem is that data can be locked in silos that can be difficult to access and this can slow down results or cause inconsistent results with too many copies of data. Another challenge is budget and energy constrains that must be contained with growing amounts of resources required for AI projects. Finally, a business must ensure that their data is trustworthy and secure as compromised or inaccessible data data leads loss of business or incorrect results that can dangerously impact the business. 

IBM Storage Scale and Scale System are designed to break through data barriers and challenges that leaders face when deploying an AI infrastructure.

 Our value is simple and centered on 4 essential data services:

• Access services with AI-Direct Access to expedite AI workloads with faster results
• Caching or core services with AI-Direct Accelerator and AI-Direct Mobility that enable flexibility with global data access
• Data management services with AI-Direct Orchestrator  that support green and lower cost initiatives
• Resiliency services with AI-Direct Resiliency that  safeguard  and create a trusted AI platform

Our story is not just about another storage product but is about delivering AI innovation into your existing infrastructure to drive more value throughout the organization. From Nvidia to watsonx and to any AI workload, Storage Scale is engineered to provide faster access to more data with less resources all wrapped with cyber enhanced data resiliency. We are AI Storage for business with performance for hybrid cloud architectures.

Forschungszentrum Jülich will be home to Europe’s 1st exascale computer – called JUPITER. It will be the first in Europe to surpass the threshold of 1 trillion (1 followed by 18 zeros) calculations/second with 6,000 nodes using Nvidia Grace Hopper GPUs (GH200) and required the fastest and most efficient storage to optimize this huge investment.

Resources:
IBM Storage Scale System for data and AI  
Webcast : Accelerating AI and IBM watsonx with Nvidia GPU access and hybrid cloud by design   
JUPITER technical overview, a deep dive into JUPITERs building blocks