Modelling layer 2 and layer 3 device bandwidths using B-node theory

Modern computer networks contain an amalgamation of devices and technologies, with the performance exhibited by each central to digital communications. Varieties of methods exist to measure and/or predict these performance characteristics. "Rule-of-Thumb" is subjective and based on prior experience, typically offering little mathematical rigour. Benchmarks use different scales and units, with comparative results possibly requiring further interpretation. Stochastic modelling uses complex mathematics which can be problematic and difficult to understand and conceptualise to the typical network administrator. As such, the specific technique employed depends on the problem domain and the cost of getting it wrong.Bandwidth-Nodes (B-Nodes) are a high-level bandwidth-centric abstraction used to de-couple and control the complexity of a particular technology from the underlying implementation. Devices and/or technologies can be modelled as an individual node or as a collection of nodes, describing the overall function and interactions between both the sub-systems and the operating environment.This paper uses a simple, common measurement method to calculate the theoretical maximum bandwidth of a single and/or collection of B-Nodes. It demonstrates that the efficiency of B-Nodes can be decomposed and shown as a product of all efficiencies contained within that node. Sub-optimal operation and device efficiency and its effect on bandwidth is also introduced. These are empirically validated and incorporated into the B-Node formula, allowing the bandwidth of a network to be calculated to a first approximation for a variety of devices and technologies. Hence, the anticipated performance of a network given a technical specification can be easily and quickly determined.