The economics of statistical multiplexing for broadband networks

We assess the overall economic benefits of good traffic management practices by quantifying the impacts of statistical multiplexing on broadband networking costs. We consider realistic baseline networking demand scenarios derived from actual data, vary allowable multiplexing gains and use an automated tool to design a near-optimal network to support the demand. Then, using current costing information, we estimate the cost savings due to statistical multiplexing under a variety of other scenarios: varying demands; differing communities of interest; with/without edge switching; and with increased demand uncertainties. We observe that setting multiplexing gains too high can result in significant service risks without appreciable cost reduction, while over-engineering can be prohibitively expensive. In all the cases studied here, we observe that there is considerable potential for reducing networking costs (between 10%-40%) even with modest levels of multiplexing gain, which are currently feasible using appropriate methods. More significantly, our analysis suggests that there is not much scope for further reductions in capacity costs through sophisticated traffic management strategies (e.g., shaping and policing, ABR), which therefore need to be justified on other considerations, such as increased robustness or simplified operations.