Maximum Size Matching Method Realized by Hopfield Neural Network for ATM cell scheduling

Abstract

Maximum size matching (MSM) cell scheduling in the ATM switching fabrics (ASF) with virtual output queuing (VOQ) is complied by a new Hopfield neural network (HNN). A new energy function of HNN is proposed to correspond to the cell scheduling rules of MSM. The new HNN scheduling algorithm employs all cells updated synchronously and realizes the MSM of the cells with global optimal between input queues and output queues in a single time slot. This difference from the iSLIP and FIRM algorithms makes it have better performances. The simulation results of 8*8 and 16*16' ASF, compared with iSLIP and FIRM methods, demonstrate that the proposed algorithm has faster convergence speed and reduces the mean delay greatly without the throughput degradation.