Welcome to Next Platform TV for today, July 7, 2020.
We are featuring an in-depth interview with Dr. Satoshi Matsuoka, director of the RIKEN Center for Computational Science in Japan and professor of computer science at the Tokyo Institute of Technology. Matsuoka has been involved in the design of so many different supercomputers it is hard to keep track of them all, but the most important one is the Fugaku supercomputer at RIKEN has just won the Triple Crown in supercomputing, with the top performance on the HPL, HPCG, and HPL-AI benchmarks, or the Grand Slam if you count the Graph500 test as well. Like its K supercomputer predecessor, it is an all-CPU system with fat vector engines and a torus interconnect, and we expect it be around for a long time like K. But will Fugaku be the last of its kind? We ask that and many more questions.
Keeping on the HPC front, we talk about high performance, high efficiency exascale computing with Dr. Giovanni Agosta, one of the creators of the RECIPE framework for energy optimization around thermal hotspots on today’s largest supercomputers.
The program turns to Dr. David Bader, a pioneer in HPC and later, big data/analytics at scale where we discuss where AI/ML fit into both areas and what’s on the horizon for both systems and software, especially as data volumes and complexity continue to grow.
Our focus shifts to quantum computing for scientific computing later in the show with Dr. Matthias Moller who describes offload and programming frameworks for hybrid HPC and quantum computing workloads on the horizon.
We end the program Gary Smerdon, CEO at TidalScale, about the new generation of virtualization and composability that is now available in the datacenter, and representing something that was not invented by the hyperscalers (for once). But in the end, given the benefits that being able to glue machines together into shared memory systems as well as slice them up into virtual machine partitions – or run bare metal-ish with Kubernetes on top – they might either use TidalScale or invent something very much like it. Cloud builders and large enterprises will similarly be very keen on this new wave of composability, which goes far beyond virtualizing storage and networking and compute on a single machine.
Cheat Sheet (Timestamps)
1:47 – Dr. Satoshi Matsuoka (Fugaku system)
19:57 – Dr. David Bader (HPC to Big Data to AI, future directions)
31:07 – Dr. Matthias Moller (Quantum programming framework)
37:38 – Dr. Giovanni Agosta (RECIPE framework for exascale computing)
41:53 – Gary Smerdon (Virtualization and hyperconverged innovations)