File and Object Storage

 View Only

Modernizing AI infrastructure for healthcare research with NVIDIA, IBM Cloud Paks with an IBM Data Fabric and IBM Spectrum Scale with IBM's Global Data Platform

By David Wohlford posted Wed December 22, 2021 12:18 PM


AI workloads are accelerating but accessing global data and scaling AI remains a challenge.  Customers are racing to become more agile and modernize applications for AI to either gain a competitive advantage or in the case of customers conducting research they also need better collaboration along with faster results for their workloads.

The challenges that prevent organizations from achieving the levels of business and operational agility they need to foster innovation and remain competitive in today’s markets should be obvious.   You hear them mentioned all the time – growth, speed, variety, and veracity of data.  But there is so much more that is creating a sense of urgency for organizations - it’s the underlying services that  IT infrastructure delivers, consistently, to enable that agility and enhance workforce productivity without compromising business requirements. To highlight a few that are affecting the way our customers are doing business today, consider the following:

1.The growth of data is constant – leaving organization’s continually facing the challenges of aggregating, managing, governing, and creating new value from their data
2. Organizations don’t often realize there is a problem with their data in advance of a situation.  Reacting creates in-the-moment data management solutions that fail to scale as the business grows, applications change, and new data formats are adopted
3.A lack of processes, data management systems, and inadequate data strategies contribute toward data quality issues and untrustworthy data sets.
4. Lack of data ownership is a key shortfall for most organizations.  This fragmentation is often across different departments and stakeholders, versus the business as a whole.
5.Many organizations are unable to invoke enough support to improve their data culture.  This is often driven by a lack of knowledge or skill set on the proper techniques to Manage data effectively.
6.Lastly, reactive approaches, incomplete processes, fragmented data ownership, and insufficient skills are catalysts that foster single-purpose solution adoption that ultimately traps data in technology silos that ultimately increase compliance and productivity risks for the business

IBM Global Data Platform

Customers need a new approach and a new modern infrastructure to their challenges if they are going to get ahead and reach the full potential of AI and the hybrid cloud.

That's why IBM created the Global Data Platform.

The Global Data Platform provides the foundation to simultaneously connect multiple data sources using multiple data interfaces from edge to core to cloud.  We provide “faster access” with “simultaneous parallel data paths anywhere, everywhere and for anyone.

Customer love our results. “We did testing with AI and machine learning workloads with Spectrum Scale and [it] outperformed all other file systems that we tested against.”   Senior system administrator (of IBM Spectrum Scale)

 “We’ve managed to unify the entirety of the data storage where we’re putting everything into one holistic fabric.” CTO of research and advanced computing (user of IBM Spectrum Scale)

Major US Healthcare research success story

A multi-billion dollar healthcare research organization wanted faster results for researchers so they could solve major healthcare issues faster and with more accuracy. They also wanted to provide more collaboration of data so more information could be shared and leveraged across the organization as silos of data had been slowing research progress and also making it more difficult for IT to manage multiple complex environments.

The customer decided on a solution that is  rooted with IBM's Global Data Platform and IBM Spectrum Scale using over 1PB of IBM ESS 3200's that provides up to 80GB/s per scalable node.  The solution provides multiple NVIDIA A100 GPU enhanced DGX AI compute nodes that are connected to a remote Red Hat OpenShift environment that will provide additional AI analysis using the IBM Data Fabric and IBM Cloud Pak for Data.  The solution is all grounded in a single data platform that simplifies the infrastructure and lowers the cost of data management and data capacity by eliminating the cost of creating multiple copies of data across the organization.

The solution provides three key value points for the research team:

OpenShift application modernization provides more application agility as customer builds a global environment to promote more collaboration and new modern containerized applications for faster development and easier application deployment environments. 

Global Data Platform provides a fast and easy way for DevOps and data scientist to move more quickly with application modernization and leverage global resources including the public cloud and hybrid cloud data at the same time helping to lower the cost of IT resources.

NVIDIA AI and IBM Cloud Pak for Data tools providing modern analysis tools with high-performance access to a global data source as customer is expecting to expand quickly their data lake in a way that integrates with new modern workloads and AI enabled solutions.

What was the result for the customer?  

The customer has now enabled a faster path to data and application modernization at a cost savings over multiple data silos with more collaboration and faster results for their critical workloads.

To learn more about IBM Storage with NVIDIA visit our new NVIDIA solutions page.

To learn more about IBM Storage with IBM Data Fabric and IBM Cloud Paks or other AI solutions visit our AI solutions page.