The increasing demand for air traffic in the last years has led to a h
eavier use of airports and airways, while their capacities have not gr
own accordingly. The main drawback of this phenomenon is a situation o
f congestion in air traffic networks that produces departure delays an
d queues before landing, causes large losses to air companies and affe
cts air traffic safety. A way of reducing congestion is to adopt a Gro
und Holding policy, i.e., to hold on the ground a limited number of fl
ights before departure in order to avoid as much as possible airborne
delay. The Ground Holding Problem (GHP) is that of determining a way o
f distributing delays to the flights in such a way as to minimize the
overall cost of the delays (both on the ground and in the air). The im
portance of Ground Holding policies is well recognized and optimizatio
n models have been proposed; unfortunately, it is very difficult to ha
ve real, or ''even realistic'', GHP instances to evaluate the quality
of one procedure over the others. Given this lack of GHP instances, in
this paper we introduce 32 new test cases, up to 5000 flights on a ne
twork of 10 airports, in which congestion is caused by insufficient ca
pacity in arrival airports. These instances (made accessible via ftp)
are solved to computationally compare a new heuristic algorithm with b
oth a previous heuristic and an exact algorithm. The new algorithm we
propose is based on ''priority rules'', where the flight priority is c
omputed as a cost function.