By Terry D. Shaw, Ph. D., CableLabs
The CableLabs Bandwidth Modeling and Management project addresses the use, management, performance simulation, and network economics of cable high-speed data systems. On this project, we have analyzed consumer usage patterns based upon data collected on live cable-based high-speed data systems as well as network simulations. Usage data on live cable networks indicate that traffic flows are fairly predictable over DOCSIS™ networks. Some of the primary patterns emerging include: • Skewed distribution of bandwidth consumption. As a general rule in many systems, 30% of the subscribers consume about 60% of the data. • Students drive seasonal characteristics. System usage is appreciably higher during local school holidays and vacation periods. • Usage rapidly evolving. In the data we collected, the average per capita use of data shows steady growth over time in both the upstream and the downstream, with upstream use growing more rapidly. This results in the average downstream/upstream symmetry ratio trending downwards, with a significant number of subscribers using more upstream than downstream data. These observations indicate the importance of deploying DOCSIS 1.1 in order to meet this increasing demand for upstream capacity. Preliminary simulation results indicate DOCSIS 1.1 will enable a significant increase in upstream system carrying capacity over the already substantial capacity of DOCSIS 1.0. These simulations indicate that DOCSIS 1.1 will allow almost 20% more upstream capacity than DOCSIS 1.0. Simulations have also been used to study the characteristics of specific types of traffic found on cable networks. For example, simulations of peer-to-peer applications indicate that one user without rate limits can consume up to 25% of upstream capacity with a usage pattern resembling a very high speed constant bit-rate application. These simulation results highlight the potential benefits for cable operators to manage their bandwidth using tools such as rate limits,