Supercomputing 2018 Offers Insights to Storage for AI and ML

By Greg Kincade - 2018-12-17

Artificial Intelligence and Machine Learning (AI/ML) have incredible potential for driving dramatic improvements in data analytics. Traditionally, AI/ML has been leveraged heavily by academia and government to generate advances in scientific analysis of various topics such as weather, seismic, social sciences, and astronomy, but that is starting to change as commercial enterprises – large and small – look at ways to leverage these deep insights into their own customers. I got a chance to attend Supercomputing 2018 (SC’18) in Dallas a couple of weeks ago … not just as a customer, but as a vendor of advanced technologies that could potentially benefit the AI/ML world.

My primary attendance goals were as follows:

  1. I wanted to gain a better understanding of how AI/ML and Deep Learning (DL) are impacting the HPC world.
  2. I wanted to better understand how Micron can leverage its partners to provide better suited products for these environments.
  3. I wanted to take the time to meet other SC’18 attendees and vendors to better understand what their ideal storage offerings would look like for HPC and AI/ML/DL.

I came away with several impressions from SC18:

  • I was surprised by the number of booths sponsored by academia (I am so used to industry events being driven by technology vendors). It seems that because of the historic prevalence of AI/ML in academia and government uses, it is important to have these entities highlight their current research.
  • AI/ML is very popular and doesn’t seem to be slowing down. There was clearly interest in using AI/ML/DL for scientific research. Deep Learning is especially starting to change the way AI systems are designed and used.
  • I learned that the world’s largest supercomputer was a $200M IBM “Summit” system with 28,000 GPUs and 9,000 CPUs!
  • Micron can help in the world of AI/ML. In fact, we had enthusiastic participation by attendees in our booth. Clearly storage is a key factor in AI/ML.
  • There is a growing interest in AI/ML in the non-public, commercial world. Companies are collecting tons of data from websites, social media and intelligent devices, and that data can provide a deeper understanding of their customers.

SC '18

All of this points to a single conclusion in my mind: AI/ML applications are key to some very specific commercial use cases:

  1. Prediction of demand for goods and services such as retail, banking and other ecommerce applications
  2. Security … especially in the world of regulated banking and healthcare industries. Being able to identify immediately when there is a security breach or a potential security breach can save billions of dollars annually.
  3. Preventative maintenance of complex mechanical environments by predicting failure before it happens can save money and lives.

Each of these use cases depends on extremely rapid computational analysis that demands instantaneous access to petabytes of data at any given instant. Micron’s advanced technologies in memory and storage technologies, such as GDDR6 memory, 3D Xpoint non-volatile memory and storage as well as advanced flash storage, blow the doors off traditional, legacy hard-disk drive deployments. As presented at the Micron Insight event last October, Micron is dedicated to driving innovative solutions in AI/ML/DL. I came away from SC18 excited by the opportunities before us … as I am sure you are as well.

Stay up to date by following us on Twitter @MicronStorage and connect with us on Linkedin.

Amit Gattani

Greg Kincade

Greg Kincade is a Sr. Ecosystem Enablement Program Manager for the Micron Storage Business Unit. His primary focus is working cross functionally with external collaborators and internal stakeholders to define differentiated solutions that help our customers dramatically increase the performance and efficiency of their storage infrastructure.