Data transfer from an application at a node of a network to a queue for a data channel is disabled when the amount of queued data in the queue is greater than a first limit and enabled when the amount of queued data in the queue is less than a second limit. The first queue limit preferably is increased when data transfer is disabled and the amount of queued data in the queue is less than the buffer storage capacity and decreased when data transfer is enabled and the data flow rate of the data channel is decreased. The second queue limit preferably is increased when data transfer is enabled and the amount of queued data in the queue is less than a predetermined data channel starvation threshold and decreased if, after data transfer from the application to the queue becomes enabled and an amount of queued data the same as the second queue limit has been communicated from the queue to the data channel, the amount of queued data in the queue is greater than a predetermined threshold. Preferably, the limits include an integral multiple of the predetermined amount of data in a burst, and are increased and decreased by an integral multiple of the predetermined amount of data in a burst. In response to a decrease in the available global storage capacity, the first and second queue limits may be reduced by first and second predetermined amounts, respectively, the first and second predetermined amounts preferably being the same.