General Method for Finding the Most Economical Distributed Router Architecture

In this work we present a novel method to determine the optimal parameters of a router architecture when certain router performance constraints are given. The total financial expense, or cost, is the optimality criterion. We introduce a general, essentially distributed, router architecture model, consisting of locally or remotely located forwarding engines or processing units gathered around a switch of variable speed. Given the following constraints: number of inputs, maximum line interface bandwidth, and maximum packet delay in a router, the presented method finds the optimal amount and distribution of processing power among the various available processing units and the optimal parameters for the switching element. The optimization employs an estimated market-based cost function per element and finds the most economical system solution. The results show that the optimal solutions gather around two extreme points of the solution space, distinguishable by the distribution of the processing power mass and corresponding switch speed. We discuss when, depending on the customer input, one or the other solution is appropriate.