Parallel Local Search and Job Shop Scheduling

We discuss parallel local search approaches to the job shop scheduling problem, based on edge reversal neighborhoods. Speed-up is achieved by parallel exploration of neighborhoods and parallel computation of the P longest paths in a disjunctive graph. A complexity analysis shows that the resulting parallel local search algorithm has a speed-up of O(P/ log log P) on a PRAM machine with P processors, and a speed-up of O(P/ log P) on a distributed-memory MIMD machine Furthermore, we show that the problem to verify local optimality with respect to the 1-opt neighborhood for the job shop scheduling problem is in.NC.