The simulation of the non-linear dynamic behavior of distributed routing networks using DECSIM

DECSIM, a logic hardware simulator, was used to model a router. Various load conditions were simulated on a mesh of 25 and 400 routers to characterize the network behavior both in the linear and in the nonlinear regions. Different scheduling methods were used to characterize the systems´ bandwidth. Simulation results are presented and compared to the nonlinear behavior of natural systems. For the present routing algorithm the deterministic and random scheduling schemes did not show appreciable differences in network message routing capacity prior to saturation. Fundamental congestion problems occurred in both cases under similar load conditions. The random behavior of the coin flip was shown to introduce enough variability in the system to allow some extra routing capacity once saturation was entered. The flexibility of a hardware simulator for this type of work was shown. The ability to program and monitor network activity in an interactive or batch environment makes this a flexible approach. The simulations presented confirm that this router mesh exhibits complex behavior that is analogous to that seen in natural systems