I should have done CFD work. Write some code and let it run for about a week, check your answers and repeat.
Yup. Tons of it is written in Fortran.
That was the sweet part of being an SA on SuperClusters. We couldn't touch shit while their code was running so it lead for a lot of days on the X and playing Sauerbraten in the office. Then again, there was a lot of times that the I/O operations were so huge that it would totally hose the file system and crash all the NFS servers. We would have to reboot and that caused us to lose hardware or nodes would just be like "Fuck it, I ain't doing this shit anymore." and we would have to start taking a look at why stuff was corrupted.
It also sucked when he started tinkering with his source code. He would forget to compile stuff with drivers for the interconnect and his shit wouldn't run. When we told him it was something to do with the way he complied and was his source code he would get livid and blame it on the hardware.
"Why is the system broke and my jobs not running?"
"Did you change your code?"
"No!"
"Nothing should be wrong then as we haven't introduced any config or hardware changes. Let me investigate and sift through some logs and what not, hang on just one second."
(cd /Volumes/Users/home/customer's dir/source code dir/current source code running dir/)
(ls -lat)
"Are you sure you didn't change your code sir"
"Yes! I am positive!"
"Well someone logged into your account and changed it b/c it was changed at 6:45 AM this morning."
"Oh, yeah I forgot about that, I did change some things this morning. Let me call you back."
He would hang up, make changes and recompile and I wouldn't hear from him for a while.
He had plenty of money to spend at one time and we got to test a lot of cool HW before it was released to the public. Rack, stack, cable, and test HW to get him the most cost effective solution was awesome. He ran out of money.
This Information Assurance thingy sucks pretty bad compared to it.