Thinking about specific settings on WorldBuilding.SE, it turns out I need help from people who are more in touch with CS and Clusters. For those who curious about how it started, backstory here, relevant details will be provided in the body of the Q. A note, despite that joking setting, it has some bearing in reality as a concept.
Motorola 68000 (specs: 8MHz, 1.5 Watt, 16/32bit, 16MB address space, ~100'000 FLOPs)
if we take old hardware as a calculating node and arrange about 19e15 calculating nodes of it in some sensible way, connect them onboard with 1MBit lanes, networks, etc
On paper, we may have a supercomputer, with a peak performance of 1'922'654 PetaFLOPs.
I assume, maybe incorrectly, that most or any task run in BOINC can be run on that cluster as well.
Highly compartmentalized tasks may run as well, which I hardly can name any besides hashing, some cryptography, searching.
ping to the system is 5 seconds, it isn't on the planet, so not that many realtime applications.
The problem I have is to assigns some value for the setup - a value for potential uses of that setup for modern practical applications.
Which CS problem solving can be run on such hardware and earn money?
Or is it more on paper and modern-day deep learning, big data, etc aren't fit for that cluster of low-performing nodes?