1,359
edits
Changes
Jump to navigation
Jump to search
Line 102:
Line 102: − + + + + + + + + + + + + + +
→Running the MIR Solver/Scheduler
== Running the MIR Solver/Scheduler ==
== Running the MIR Solver/Scheduler ==
=== Questions about MIR ===
=== Questions about MIR ===
*
*Does the head node talk with the slave nodes?
*Do the result files keep track of the best solutions?
*How is the job divided amongst the nodes?
=== Answers to the MIR Questions ===
RESULTS (thus far):
The head node does not exchange answers with the other nodes. The
BestCostFunktion file only contains the head node's information.
justification: the minimum value of the head node's costfunction was the minimum found in the BCF file. While other nodes had found lower solutions. I do not have proof of the 22991 value, but I do have proof of solutions in the 20k's.
As far as I can tell now, the nodes all go through runs in unison. That is
they each heat up cool down 10 times.
justification: A visual inspection of each of the nodes' cost functions shows less and less hill climbing as step number progresses (and temperature decreases). Since the number of this Cost Function density is periodic in step number, equalling about 7 in total (out of ten). The 7 is a result of the information from each node being taken before the end of the run. (NOTE: the claim about temperature versus the costfunction will need to be tested.)