Flow-level models for capacity planning and management in interference-coupled wireless data networks

In 4G cellular networks, both the adaptation of data rates to current interference conditions due to adaptive modulation and coding as well as a frequency reuse of one mandate precise techniques to estimate cell capacities and cell loads in order to accurately predict the quality of service delivered to end users. Such estimation happens ideally already during the network planning phase and is further required for self-optimization at runtime. Classic flow-level techniques to estimate cell loads, capacities, and related quality of service metrics assume static and worst case interference, which is analytically simple, but may produce considerable errors and lead to disadvantageous planning and optimization results. Appropriate models where individual cells are coupled through interference are rendered analytically intractable. This article first introduces basic flow-level modeling techniques and then reviews recent results in the field of flow-level network models, which allow the actual loads and capacities in interference- coupled wireless networks to be bound and closely approximated. We discuss trade-offs between accuracy and numerical complexity of different techniques and identify a model based on the notion of average interference as the most practically relevant. Simulation results for a large scenario based on a real network illustrate its applicability to practical network planning.