This paper reports on research into novel optimisation schemes for large-scale distributed computing environments that will enable data analysis and knowledge acquisition in the course of optimisation. The scheme incorporates concepts from the Simulated Annealing search strategy in order to ensure robustness. In contrast to Simulated Annealing, which is a sequential optimisation algorithm, the proposed optimisation scheme consists of a number of solution pools, each of which is associated with a system temperature which defines solution quality within the pool. The solutions in these pools are generated by performing constant temperature Markov processes on existing solutions in these pools. As the individual Markov processes are independent they can be completed in large-scale distributed, computing environments, constantly producing new solutions which are stored in a central database. During the optimisation, the solutions are regularly reassigned to pools according to their performance relative to the other solutions that have been generated such that the solution quality improves towards the pool associated with the lowest temperature. This final pool accumulates the set of optimal solutions during the optimisation. The solutions of all pools are stored in a central database from which knowledge about the importance of individual solution features can be extracted in the context of the systems performance.
- distributed computing
- knowledge acquisition
ASJC Scopus subject areas
- Chemical Engineering(all)
- Computer Science Applications