Broadband Capacity Metrics
One of the challenges in developing "Erlang Formulas" for broadband networks has been to develop theoretically consistent and empirically observable quantitative measurements of both demand and capacity for the network. Demand can be difficult to objectively measure because it can be highly dynamic, or "fractal"/chaotic in nature -- particularly for interactive HTTP traffic. TCP-based flows competing for a broadband bottleneck's bandwidth require a "new measure" for this capacity and their associated demand.
In the early 1990s a group of researches discovered the self-similar (or "fractal") nature of network traffic and a great deal of theoretical and practical work has been performed to develop new ways of measuring demand and capacity (see Self-Similar Network Traffic: An Overview). Meanwhile, empirical evidence has shown that for backbone networks, the fractal properties of the traffic are not relevant for gross over-provisioning to insure quality of service (QoS) as some of the theoretical "fractal" models had previously assumed. Therefore, new measures appropriate for provisioning and insuring stream blocking probabilities for Internet video have to begin with managing the capacity of broadband networks. These regions of the Internet are the most highly and easily congested locations today.
Unfortunately, from a business standpoint the first key step in articulating any solution to the problem must involve clearly specifying the capacity that is being sold to an end-customer, or broadband user. InterStream quantitively and objectively specifies this capacity through establishment of "mediation policies". These policies may be used as a basis setting Terms of Service and Acceptable Usage Policies for broadband providers. They also allow consumers to objectively compare service levels from competing broadband providers. Through use of designated TCP implementations, the streaming demand may be clearly quantified against mediated capacity. In addition, when ISTP is operating in "stealth" mode on broadband networks, the bottlenecks can be effectively mediated by the TCP implementation. In effect, an Erlang formula may be offered to consumers and broadband providers to objectively measure network capacity. InterStream media industry streamers may use the formulas to set minimum quality standard for use of the association's service mark to insure their Internet television service offerings are supported by all ISPs.
Measuring the demand, "mediation" of the capacity, or core bandwidth and traffic management must be offered to ISPs. Our goal is to normalize bandwidth demand through the use of a designated TCP implementation. This was the premise behind creating a new form of TCP congestion control. By combing these two aspects of capacity and demand, a standardized quantitative measures may be used by broadband providers, consumers and media companies in the InterStream system. Formally, the capacity of a specific broadband connection may be specified via the "Mediation Policy" offered by an InterStream Broadband Service Provider. Streaming bandwidth can be expressed in simple, bit per second form, if a certified version of InterStream's TCP congestion control algorithms are used as demand for that capacity.