It's a fun article, but it actually drives me to a question: is it possible (by which I mean remotely feasible) to design a user experience from the ground up around a massively parallel system?
Let's say you had some low speed hardware to support a beefy GPU in a case, could a manageable operating system be built that wouldn't spend all of it's time in IO-wait? I like the idea and challenge in designing this way, but I don't see a way forward that doesn't just fall back on the architecture of core computational unit with specialized auxiliary units quite rapidly.
That gives me even less faith in the quality of findings.
Having seen the pathetic state of literature review in the various engineering fields, I cannot expect it to be better in meteorological fields. No doubt there is a certain subset of effective and critical reviewers, but I expect dogma, exciting findings, and name recognition still rule the day.
The most the volume of politics in engineering where our work is not overtly political to those outside outfield is astounding. I don't buy that in a field with so many non-expert pressures applied the review process escapes unadulterated.
A friend of mine works in IT for district doing this. They have under a dozen IT staff serving five k-12 schools. In the past he worked in a much larger distinct with a Dell based laptop program.
One of the reasons they love the Chrome system rather than a more traditional install (be it Linux or Windows), is when some clueless student comes in saying: "it broken," the staff can wipe the device clean in a second. The student can log back into the Google services to bring their account back, and they minimize undue downtime to repair software problems created by the students.
Could you get something similar going under an alternative distribution, sure, but with this they're is no labor/cost overhead associated with maintaining the server systems and software.
Does anyone else see a wonderful watchdog project for formally/publicly identifying inept instructors and professors coming out of this?Or is that just foolish optimism on my part?
Let's say you had some low speed hardware to support a beefy GPU in a case, could a manageable operating system be built that wouldn't spend all of it's time in IO-wait? I like the idea and challenge in designing this way, but I don't see a way forward that doesn't just fall back on the architecture of core computational unit with specialized auxiliary units quite rapidly.