Internet access and capacity planning: Quantifying relationships between usage, capacity, and blocking

The blocking probability is commonly used to evaluate the quality-of-service (QoS) offered by dialup Internet Service Providers (ISPs). Unfortunately, dialup ISPs employ no blocking standards or thresholds. This raises questions as to exactly how providers measure and evaluate QoS with respect to network access and how they make capacity planning decisions. Many ISPs use the User-to-Modem Ratio (UMR) as a surrogate measure for service quality, but there are no well defined direct relationships between the UMR and the blocking probability. In this paper, relationships between the UMR and other well known traffic variables and the blocking probability are quantified, and the usefulness of the UMR as a standalone QoS metric is explored. Blocking probabilities are estimated for a range of UMRs via simulation from a range of demand scenarios developed from observed data. A predictive regression model is introduced to quantify the affects of independent predictor variables on blocking. Results show that the UMR is not a useful standalone QoS metric and traditionally favored UMRs such as 10:1 cannot be associated with "high quality service" in a de facto manner.