Featured Image (top)
Multicore World 2019: (L-R) Nicolás Erdödy (Open Parallel, Moderator), Satoshi Matsuoka (RIKEN), James Ang (PNNL), Geoffrey C. Fox (Indiana University), Jeffrey S. Vetter (ORNL), Andrew Ensor (AUT University)
Updated 20 January 2020
Panels are one of the most lively parts of Multicore World. Each topic is both challenging and broad enough to entice speakers and audience to openly debate towards mutual enlightenment. The moderator triggers the debate presenting a few questions as guidelines and the panel and audience freely discuss around the topic. No slides are used. Controversial opinions are encouraged: it’s a debate after all!
Updated 20 January 2020
Panel 1 – Date (TBC)
HPC and Smart Systems
HPC tends to use brute force to produce solutions. Solve the equations and the solution is in your hands. That is however not sufficient for AI and ML (“Smart Systems”).
i) Can HPC benefit from this and vice versa?
ii) Does the HPC ecosystem change under the avalanche of different accelerators and other types of special hardware currently hitting the market?
iii) Is technology independence already a thing from the past?
iv) Does a TCO change trigger a shift in the HPC demographics, or will everything be Obscured by Clouds?
v) Are we going to face the “Cloud Supermarket Dilemma”, where our choice of systems is restricted to what is offered on the shelves of a specific cloud vendor we picked? What if we want something less mainstream?
Panel 2 – Date (TBC)
Legacy, Heterogeneity and Integration
Every system architect and engineer knows that we have to deal with this, but thanks to the users, they’re always behind. The reason is that users keep on re-inventing wheels, not necessarily round, as opposed to improve the existing infrastructure. Those new languages and tools often do not leverage the system features and trigger the need for new ones. That takes several years to implement and meanwhile the new kid on the block is there already. And that is how the spiral continues to go downward.
i) How do we avoid this ongoing issue to continue in the exascale era?
ii) Is there a reasonable way to future-proof our ever growing systems? And is this a purely technical issue or a technology management issue?
iii) And what about security between legacy and new systems? Aren’t we just creating conditions for “Black Swans” to emerge?
iv) Should we get paranoid and apocalyptic or just wait till the next quarterly results?
Panel 3 – Date (TBC)
Exascale at the Edge
Edge computing is the “next new thing” but…
i) Is this actually something radically new? What’s the “winning component” in the solution? I/O? Cybersecurity? New languages? “Minimalist” HW & SW?
ii) Would mini datacenters near to where the data is produced simply solve the challenge?
iii) Could they be combined to operate at exascale level on a system of systems?
iv) Should the HPC community be involved with this, leveraging their skills?
Updated 11 February 2019
Panel 1 – Tuesday 12 February 2019 – 3:30 pm – 4:10 pm
The S in IoT is for Security.
And the H?
Given the lightbulb in your toilet is connected to the Internet, security has to be the responsibility of the vendors, not the end user: Every IoT product should have a security emergency scenario built-in. Should we be more worried about the impact of cyber attacks because of IoT?
More down to earth: Every programmer needs to know the basics of processor and systems architectures but…Where is IoT in HPC?. Someone would claim that it is non-existent and that would be a shame.
Related to that, does the HPC community care about fault tolerance in software?
And let’s think through primary sector applications: i.e. Precision Agriculture, Forestry, and Fisheries. What new problems does precision primary industry bring to computing? What new opportunities do HPC and IoT bring to primary industry? How do we bridge the gap between the large amounts of data we could record and the people who need to respond to it?
Beckman, Moir, Schramm, Seager, Vincent
Panel 2 – Wednesday 13 February 2019 – 3:30 pm – 4:10 pm
Strategies to grow an economy with HPC
This panel should explore opportunities and models for midsized countries like Australia and New Zealand.
Can you give examples where high-end HPC projects and/or Multicore R&D have been adjusted down to fit the smaller sized companies and economies?
Would you consider that “the funding cycle” makes scientists lazy performance-wise? i.e. High-end projects accept too many customised proprietary “solutions” by vendors. Are these extremely high-end HPC projects a good use of public funds? Are the software investments made for the long term, or do we code for today, not tomorrow?
Do you envisage potential gaps in mature markets -such as Finite Element Modelling, brought by new Exascale architectures?
What economic growth strategies would you suggest that apply next gen computing to the primary sector, while processing a systems conversion / transformation from “the first” generation of computing (~ 60-70 years of IT) into the Exascale era? and how would this impact New Zealand?
Ang, Ensor, Fox, Matsuoka, Vetter (plus contributions from Seager 😉
Panel 3 – Thursday 14 February 2019 – 11:35 am – 12:10 pm
Multicore, IoT and the OS – Close to the Edge?
TOP500 supercomputers run Linux. Embedded systems run (sort of) Linux. Mobile phones run Android (based on Linux). Have we reached some kind of optimum? Or is this just a historical/commercial accident? Can we reach exascale size with IoT data? What do we actually need from exascale operating systems? And can an exascale OS run well on IoT at all? Do we even know what questions to ask?
And…what could we do for/from New Zealand?
Blanchard, Foster, Hendel, van der Pas, Stark
These were the panels at Multicore World 2018.