Ayar and Yaman (2012) use a real world data set with 34 nodes
and 167 services, and generate a random network of 66 nodes and
1200 arcs, and test them for 400 up to 1000 commodities. Their
first conclusion is that by increasing the number of commodities,
the optimality gap decreases. Moreover, adding valid inequalities
and variable fixing strategies result in significant improvements
both in time and solution quality. All their instances are solved
to optimality within 6 min. Ayar and Yaman (2012) also investigate
variable fixing based on capacity restrictions, but in overall, it does
not bring much improvement to their previous valid inequalities
and variable fixing strategies.