Abstract
The industry is satisfying the increasing demand for wireless bandwidth by densely deploying a large number of access points which are centrally managed, e.g. enterprise WiFi networks deployed in university campuses, companies, airports etc. This small cell architecture is gaining traction in the cellular world as well, as witnessed by the direction in which 4G+ and 5G standardization is moving. Prior academic work in analyzing such large-scale wireless networks either uses oversimplified models for the physical layer, or ignores other important, real-world aspects of the problem, like MAC layer considerations, topology characteristics, and protocol overhead. On the other hand, the industry is using for deployment purposes on-site surveys and simulation tools which do not scale, cannot efficiently optimize the design of such a network, and do not explain why one design choice is better than another. In this paper we introduce a simple yet accurate analytical model which combines the realism and practicality of industrial simulation tools with the ability to scale, analyze the effect of various design parameters, and optimize the performance of real-world deployments. The model takes into account all central system parameters, including channelization, power allocation, user scheduling, load balancing, MAC, advanced PHY techniques (single and multi user MIMO as well as cooperative transmission from multiple access points), topological characteristics and protocol overhead. The accuracy of the model is verified via extensive simulations and the model is used to study a wide range of real world scenarios, providing design guidelines on the effect of various design parameters on performance.
Introduction
As the demand for higher data rates grows, the wireless industry has pivoted toward "densification"—the deployment of numerous access points (APs) in close proximity within environments like airports, stadiums, and corporate campuses. While this increases capacity, it introduces significant interference challenges.
Existing analytical approaches generally fall into two categories: academic models that often rely on oversimplified physical (PHY) layer assumptions or ignore MAC layer contention (CSMA/CA), and industrial simulation tools that are computationally expensive "black boxes." This paper bridges that gap by introducing a unified analytical model that is both mathematically rigorous and practical for large-scale network optimization.
Methodology
The proposed framework integrates channelization, power allocation, and user-AP association. The core innovation lies in two distinct modeling approaches:
MAC Layer Modeling
The model utilizes a Continuous Time Markov Chain (CTMC) to characterize how APs contend for airtime in an 802.11 environment. By representing the network as a conflict graph, the model identifies valid transmission states where no two interfering APs are active simultaneously.
As shown in the conflict graph above, nodes represent APs and edges represent interference. The CTMC allows for the calculation of limiting probabilities for each feasible state, which in turn determines the "airtime" share for each AP based on the Clear Channel Assessment (CCA) threshold.
PHY Layer Modeling
To account for advanced techniques like Multi-User MIMO (MU-MIMO) and Distributed MU-MIMO, the authors employ asymptotic formulas from random matrix theory. This approach transforms complex, stochastic fading into deterministic formulas, enabling rapid computation for large antenna arrays without sacrificing accuracy.
Evaluation Scenarios
The model was validated against industry-standard simulators (NS-2 and Monte Carlo) across four real-world topological scenarios:
- Conference Hall: An open 20m x 20m space with 200 users.
- Open-Floor Office: A 160m x 23m cubicle environment without signal-blocking walls.
- Office Building with Rooms: A corridor-based layout with significant pathloss due to walls.
- Stadium: A massive 200m x 200m scenario with 20,000 users, testing the model's scalability.
Key Insights and Results
- Scalability and Accuracy: The analytical model's per-user throughput results closely matched heavy-duty simulations while remaining computationally efficient enough to analyze a 20,000-user stadium scenario.
- The Impact of Physical Barriers: While distributed MU-MIMO (AP coordination) provides massive gains in open spaces, these benefits diminish in walled offices where natural isolation makes independent transmissions more efficient.
- Optimization Levers: Tuning the CCA threshold alongside power control was found to be a critical factor in boosting total network capacity. Furthermore, sectorization (using directional antennas) can improve performance by 3x to 7x by naturally reducing contention.
- Quantized Rates: The study emphasizes that using theoretical Gaussian rates can be misleading; real-world hardware uses discrete Modulation and Coding Schemes (MCS), which must be accounted for to ensure accurate optimization.
Conclusion and Future Work
The introduced model provides a scalable tool for optimizing next-generation wireless deployments. Future refinements will focus on the signaling overhead required for Channel State Information (CSIT) collection and the potential bottlenecks introduced by the wired backhaul in highly coordinated distributed MIMO systems.
Summary updated