flow computing systems
What if we built the ultimately scaled computer system for making discoveries? How would we do it? How would we program it?
By Mark Anderson
In this issue, I will lay out what the British would call a "programme" for the future of computing, at the largest and most advanced scales.
This work was done with many others, over several years, through a combination of private work, SNS publications, developments at Pattern Computer Inc., and CTO Design Challenges at the FiRe (Future in Review) conferences, beginning in 2015.
My interest in the issues around very high-level computing architectures and challenges was reawakened during our May FiReSide Event, which featured Larry Smarr and Bob Bishop. Bob is working on Earth simulations, and as head of the ICES Foundation and past CEO of Silicon Graphics, he is uniquely qualified to do this. Larry is working on the Pacific Research Platform (PRP), and as past executive director of the UC San Diego Qualcomm Institute / Calit2 labs, and as the original founder of the NSCA, he is uniquely qualified to discuss hyperscale computing.
In both cases, the result of last week's FiReSide conversations was to paint a fascinating picture of the application needs for a compute system that could reflect the planet, either in simulation or as a digital twin.
Larry drove the vision to the point of elucidating a realtime, self-reinforcing system for a Mirror Earth, using the PRP network as an example, which connects most of the major university systems in the US as well as select other similar installations in Tokyo, Amsterdam, and other locations. Today's tools for doing this, in his telling, rely heavily on technologies (much from Google as open source) such as containers and Kubernetes, running on relatively basic nodes such as Calit2's FIONA boxes as connectors.
It's a beautiful vision, and it works today.
But what about tomorrow?...