For example, if you start 1000 reactions at once (on XSEDE) and 100 of them have a reactant in common, they will each ask themselves "do I have finished log files for this reactant?" and decide "no, I don't, so let me start calculating it". Then you are calculating the same reactant 100 times, probably fighting over saving the log files in the same folder.
This is the sort of thing we need to anticipate and avoid.
We should do something like a "HAZOP" of our code. Think about "too much flow?" etc. at each point. or "what if...".
For example, if you start 1000 reactions at once (on XSEDE) and 100 of them have a reactant in common, they will each ask themselves "do I have finished log files for this reactant?" and decide "no, I don't, so let me start calculating it". Then you are calculating the same reactant 100 times, probably fighting over saving the log files in the same folder.
This is the sort of thing we need to anticipate and avoid.
We should do something like a "HAZOP" of our code. Think about "too much flow?" etc. at each point. or "what if...".