Co-Design for Accelerating Analytics with Today's CPU & Tomorrow's Heterogeneous Compute Landscape

Accelerating and optimizing data analytics workflows has several challenges depending on your perspective and approach. Here are three examples:

  1. Database players view them from the perspective of storage, viewing analytics workload problems as an extension of database problems
  2. Higher-level programming languages and environments such as JVM result in tradeoffs between performance and ease of programming
  3. Data analytics workflows have been split between frameworks/tools that focus on analytic computation and those that focus on data visualization

In this highly informative talk, Founder and CEO of OmniSci Todd Mostak takes us on a comprehensive tour of how to use the latest HPC techniques to simultaneously accelerate analytics SQL and data visualization—a skill the company has been honing since 2013.

Topics covered:

  •  Key lessons OmniSci has learned by re-examining the nature of data-centric workflows, including how successive generations of hardware accelerators provide opportunities and unique technical challenges
  • How these workloads can be viewed as a “co-design” problem requiring an understanding of the hardware/infrastructure characteristics and the workload patterns themselves
  • Techniques to accelerate analytic workflows by leveraging hardware optimizations at every stage of the workflow—IO acceleration to LLVM-based JIT compilation to large-scale, in-situ data visualization and efficient interfaces with ML/DL workflows

Download the software

Get Intel® oneAPI Base Toolkit, which includes many optimized tools and libraries for data analytics

Additional resources:
Intel® Optane™ DC technology for data centers

Todd Mostak, Co-Founder and CEO of OmniSci

Todd is the CEO and co-founder of OmniSci (formerly MapD Technologies). Todd originally conceived the idea for MapD while conducting graduate research at Harvard on the Arab Spring. He later joined MIT’s CSAIL in the database group as a research fellow, under the supervision of Professor Sam Madden, before founding MapD in late 2013.

Performance varies by use, configuration, and other factors. Learn more at