Conclusions
Supercomputers have recently gone through a significant transformation. They have moved from custom, power-hungry, heat-generating, difficult to program monsters into power-hungry, heat-generating, difficult to program clusters. The power-hungry, heat-generating problem is a free will choice based on the ultimate performance needed. The difficult to program part is solvable through creativity and education. Newer approaches to application languages would be an enormous help. The monster part is solvable, either through technology turns or a completely different approach to packaging. Don't be afraid to think new thoughts. The supercomputer business has a long way to go and deserves to be placed in offices.
