[Paper Review] Performance Modeling of Next-Generation Wireless Networks

Antonios Michaloliakos, Ryan Rogalin+3 more|arXiv (Cornell University)|May 1, 2014

Advanced MIMO Systems Optimization40 references3 citations

TL;DR

This paper proposes a scalable, accurate analytical model for next-generation wireless networks that integrates physical (PHY), MAC, and higher-layer parameters—including multi-user MIMO, distributed beamforming, and CSMA access—enabling performance optimization across real-world scenarios. The key contribution is a unified framework validated via simulation that reveals significant gains from distributed MU-MIMO, especially in dense deployments, while quantifying overhead tradeoffs from synchronization and calibration.

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.

Motivation & Objective

• To address the lack of scalable, accurate analytical models that capture real-world complexities in dense wireless networks.
• To bridge the gap between industrial simulation tools (which lack scalability) and oversimplified academic models (which ignore key practical factors).
• To provide a unified, analytically tractable framework that models advanced techniques like distributed MU-MIMO, CSMA MAC, and interference management across diverse topologies.
• To offer design guidelines for real-world deployments by quantifying the impact of key parameters such as channelization, power control, and CSMA threshold.
• To validate the model through extensive simulations and apply it to realistic scenarios including office buildings, stadiums, and conference halls.

Proposed method

• The model integrates physical layer techniques including SU-MIMO, MU-MIMO, and distributed MU-MIMO with coordinated beamforming across multiple APs.
• It incorporates MAC layer mechanisms such as CSMA/CA with adjustable CCA thresholds to model concurrent transmissions and interference tolerance.
• The framework accounts for topological factors like user density, AP distribution, and propagation loss, including path loss and shadowing effects.
• It models protocol overheads such as CSIT feedback, synchronization sequences, and calibration pilots required for distributed MIMO, with overheads discounted as a factor on goodput.
• The model uses a unified analytical treatment of various PHY schemes, enabling performance comparison across technologies like 802.11n/ac and future 5G/IMT-2020 systems.
• Validation is performed via extensive simulations across diverse scenarios, with results used to derive design insights and performance tradeoffs.

Experimental results

Research questions

• RQ1 How does distributed MU-MIMO compare to non-coordinated MU-MIMO in terms of spectral efficiency and goodput in dense deployments?
• RQ2 What is the optimal length of synchronization and calibration pilot sequences in distributed MU-MIMO systems to balance overhead and performance?
• RQ3 How do different channelization strategies (e.g., wideband vs. multi-channel) affect network throughput and interference in real-world topologies?
• RQ4 What is the impact of the CSMA CCA threshold on user throughput across different wireless technologies and deployment scenarios?
• RQ5 How do practical limitations like quantized MCS and residual CFOs degrade performance in distributed MIMO systems?

Key findings

• Distributed MU-MIMO provides significant spectral efficiency gains over non-coordinated MU-MIMO, especially in high-density deployments with coordinated APs.
• The optimal synchronization pilot length for a 20-AP, 200-user conference hall scenario is 6 OFDM symbols, after which gains diminish despite increased overhead.
• Each distributed MU-MIMO cluster of 5 APs incurs a minimum of 30 OFDM symbols of overhead for orthogonal synchronization, reducing effective goodput.
• Calibration pilot sequences must be repeated per data slot, with an optimal length of 4 OFDM symbols in the same scenario, adding 20 symbols of overhead per cluster.
• The model shows that non-coordinated approaches become interference-limited quickly, yielding no additional gains beyond a certain AP density, even with more APs.
• Sectorization offers substantial gains but requires expensive front-end hardware, making it less practical for typical WiFi APs compared to cellular base stations.