| 摘要 |
The invention discloses a method for maximizing overall system performance among a set of computer systems which periodically run a set of jobs. It is known that some allocations of jobs to computer systems are worse than others where "worse" may mean slower or even that certain jobs are not run. What is not known is how to achieve the best or maximized job mix. However, this invention can select the best from a set of job mixes or some solution which comes within a margin of tolerance for some theoretically ideal maximum. The invention takes a novel approach to bin packing in that the items are initially ordered with most difficult to pack items first. Then, the list of items is reordered using one of a set of percolation techniques. Random processes can also be applied to the list so that from a large set of possible orders, some order will be best relative to other orders if no solution is found within the pre-specified tolerance.
|
