Multicore World 2016 : A Summary
Submitted by lev_lafayette on Sun, 02/21/2016 - 00:55
Multicore World is a small annual international conference held in New Zealand/Aotearoa sponsored by OpenParallel. I have been fortunate enough to act an MC for all but one of the five conferences since its inception, this year also presenting a short paper on the introduction of the new HPC/Cloud hybrid at the University of Melbourne. In all there were over thirty papers and panels over the three days of the conference's duration and it is possible to identify a few main issues that were of importance to a number of speakers; the Square Kilometre Array, multicore hardware and infrastructure design, along with (somewhat unexpected) some interesting presentations on the interface between social and technological concerns.
As a major scientific endeavour, the Square Kilometre Array continues to be a significant subject as it has been for prior conferences with Andrew Ensor from Auckland University of Technology taking a lead with Markus Dolensky from the International Centre for Radio Astronomy. However this conference also discussed some very practical challenges of implementation in the developing world with Tshiamo Motshegwa from the University of Botswana explaining that country's contributions. There was also panel discussion on the subject and OpenParallel itself is the work package manager for the Software Development Environment work package for the CSP (Central Signal Processor) and also contributes to the Science Data Processor (SDP).
Multicore implementations are, of course, a central concern of a conference with the ongoing debate of the best way to deal with large dataset issues. John Gustafson of the Singapore Agency for Science, Technology and Research, provided an excellent extension of his universal numbers into computation of reals, Michael Wang from NVIDIA with a brief but exciting outline of their plans in the GPGPU space, and Geoffrey Fox of Indiana University (in a wide-ranging talk that covered many other issues) threw down the gauntlet on some of the design inefficiences in cloud computing and particularly OpenStack. The conference also had a presentation from Balazs Gerofi, from the RIKEN Advanced Institute of Computational Science, one of the top computer systems in the world.
Included in the third group of talks was Mark Moir from Oracle, elaborating on the social engineering issues of Bitcoin's technical design (without delving into the economics of this collectible hash), Nicola Gaston from the MacDiarmid Institute for Advanced Materials and Nanotechnology, on scientific communication, transparency, and ethics, and at two panel discussions which dealt primarily with the economics of building data processing centres and their contribution of such centres to scientific and technological progress as a whole.
This is, obviously, only a small selection and summary of the numerous talks and panels that occured at Multicore World 2016, but hopefully gives at least a taste of the event. As mentioned, it isn't a big conference, but nor should it be. Whilst everyone uses multicore and manycore computing systems these days, and it's absolutely critical, the pointy end of design and implementation is actually down to a small number of people. Nicolás Erdödy from OpenParallel deserves accolades for his work in organising the event, the speakers, the sponsors, and a great venue (right on Wellington harbour). Kudos are also given to the sponsors, especially Catalyst and NVIDIA for the sustenance. Oh, and a big thumbs down to the Australian Department of Immigration and Border Protection which prevented one of the speakers attending due to transit visa restrictions. This said, one can justly look forward to Multicore 2017!