Many supercomputers that utilize specialized hardware just can't take component failure. For example, on a Cray XT5, if a single system interconnect link (SeaStar) goes dead the entire system will come to a screeching halt because with SeaStar all the interconnect routes are calculated at boot and can not update during operation. In any tightly coupled system these failures are a real challenge, not just because the entire system may crash, but if users submit jobs requesting 50,000 cores but only 49,900 cores are available.

Checkpoints are necessary, but in large-scale situations they are often difficult. You usually have a walltime allocation for your job and you certainly don't want to use 20% of it writing checkpoint files to Lustre (or whatever high-performance filesystem you are utilizing). Perhaps frequent checkpointing works on smaller systems/jobs, but for a capability job on a large system you are talking about a significant block of non-computational cycles being burned.

The working theory is that the lack of a strong magentosphere on Mars has allowed the solar wind to cause much of the water that was once present to be lost to space.

--Michael

There are several journal articles and books that explore the effects of indirect neuron-neuron interactions, including glial cells.

This is all a part of a growing belief that while the "direct" interactions are the most significant, the indirect actions, such as electrical fields generated by groupings of neurons are significant in the brain's operation.
And, yes, there are connections between glial cells and neurons, they act, acording to many articles like the bases of transistors, altering the intensity of a neuron's impulse.

Also, since we are mentioning complex methods of communication within the network, the interactions of dendrites (the "outputs" of the neurons) can have suprising amounts of influence on each other in a so called dendretic network, a
minor neural environment at the end of each neuron.

Basically, as time and our knowledge of the brain
has increased, we see that building a brain in
a box by simulation isn't about big hardware alone. It is about understanding how the brain is a marvel of engineering and that we have a long way to go, no matter what catchy TV commercials say.

The following journal articles/books might find
some interest:

Koch, C., Poggio T., & Torre, V. (1983). Nonlinear interactions in a dendritic tree: Localization, timing, and role in information processing.
Proceedings National Academy Sciences USA, 80.

MacLennan, B. J. (1992a). Field Computation in the Brain (report CS-92-174). Knoxville TN: University of Tennessee, Computer Science Department.

MacLennan, B. J. (1992b). Information Processing in the Dendritic Net
(report CS-92-180). Knoxville TN: University of Tennessee, Computer Science Department.

Shepherd, G.M. (1988). Neurobiology, Second Edition.
New York NY: Oxford University Press.

As for software implications, not soon since there are not very good mathematical modeling systems to even start exploring this aspect of neural interactions. Plain old artifical neural networks (with their own variations of interest) will still be your best bet for betting on stocks.

Michael Campfield
Computer Science Department
University of Tennessee, Knoxville