Traffic grooming is the technique which combines low-bandwidth traffics into higher-bandwidth channels in order to satisfy certain constraints as to maximize the traffic throughput, minimize the connection-blocking probability, or optimize the wavelength bandwidth exploitation. A case considered in this paper is optimization of the multiplexing of customer services into bursts at the edge nodes of an optical network. This is formulated as an integer linear program and dealt with basing on the Hopfield network. The methods based on neural networks normally require a considerable delay which is hard to practice for optical data transport. This paper also proposes solutions to reduce this delay.