AWS Computing at Scale and Machine Learning

Machine Learning

Here is the continuation of my notes from the Keynote addresses at the 2017 AWS Re: Invent conference. Enjoy!

Computing at Scale

  • Growth in EC2 has been HUGE when you look at the increase in cores running in AWS
  • But GPU and FPGA core growth rates have crushed that! It has been exponential growth

Machine Learning

  • Machine Learning always seems to come up about every 5 years – then fades away again
  • This time it is “sticking” thanks in large part to the scale that the cloud offers
  • Examples:
    • Tu Simple with autonomous transport (trucking industry)
    • Matrix Analytics with tumor recognition for cancer screening
    • Clemson University and natural language learning – in the span of two hours, they had 1.1 million cores running on AWS!
  • There are three big drivers of the explosion in this technology:
    • HW acceleration capabilities thanks to technology from the likes of Nvidia (P3.16XLARGE instances)
    • Improvements in the Machine Learning frameworks
    • A rapidly growing community
      • GLUON
      • ONNX
      • NNVM and TVM

Leave a Reply

Your email address will not be published.