Airline Crew Pairing Generation in Parallel

Crew pairing generation is a time consuming part of the complex process of airline crew scheduling. We present a parallel algorithm for this problem's solution and several implementations of this algorithm on a parallel machine. Linear speedup is attained by exploitation of the combined shared/distributed memory architecture of the underlying cluster computing engine, by development of a novel communication mechanism, and by creation of application-specific load balancing algorithms. Specifically, we distribute pairing generation across both the distributed nodes of the cluster machine and the multiple processors within each cluster node. The resulting parallel computation utilizes both processes mapped to nodes and within each process, multiple threads running on the nodes' multiple processors. Since this distribution should be done in accordance with the actual workload experienced on nodes, it is performed and re-performed repeatedly throughout the execution of the parall...

• Type: Conference
• Year: 2001
• Where: PPSC
• Authors: Diego Klabjan, Karsten Schwan
• Keywords: Crew Pairing Generation, Distributed And Parallel Computing, Multiple Processors, Parallel Algorithm, PPSC 2001