The main computation was carried out on a cluster of 96 compute nodes. Each compute node has dual Intel Xeon 2.8 GHz processors with 2 GBytes of memory. The nodes are networked using switched 100 MBit ethernet.

To handle the large pools of cutting planes that were gathered during the branch-and-cut runs, we made use of a Sun Fire V480 server equipped with four 900 MHz UltraSparc III Cu processors and 16 GBytes of memory. This allowed us to run a shared-memory parallel cut-server to support the cutting-plane procedure.