Estimating the Deliverable Quality of a Fully Redundant Dispersity Routing System

Abstract

The public Internet in its current form does not provide consistently the levels of service that real-time services such as Voice over Internet Protocol (VoIP) demand. Indeed, the scope of this gap is such that quality and reliability problems are characteristic of these services. Fully redundant dispersity routing exploiting the path diversity readily available in the Internet is one approach of mitigating these quality and reliability problems. This paper presents a model for estimating the quality that may be expected from fully redundant dispersity routing systems using paths with known packet loss and loss burstiness characteristics. That model is then applied to estimate the quality that may be expected from fully redundant dispersity routing systems of 2 – 6 paths and, for contrast, to the estimated quality that may be expected from single path systems. The insights gained by this application may be useful when selecting paths for a fully redundant dispersity routing system to satisfy some quality goal. A brief study into the accuracy of the model indicates that for two paths, 50% of the estimations are within 0.05 of the simulated Mean Opinion Score (MOS), and 98% within 0.32.