[Paper Review] On the Throughput Allocation for Proportional Fairness in Multirate IEEE 802.11 DCF under General Load Conditions

F. Daneshgaran, Massimiliano Laddomada+2 more|ArXiv.org|Mar 13, 2008

Wireless Networks and Protocols12 references3 citations

TL;DR

This paper proposes a modified proportional fairness (PF) criterion for multirate IEEE 802.11 DCF that accounts for varying packet arrival rates ($\lambda_s$) and transmission rates ($R_d^s$) under general (non-saturated) load conditions. By optimizing contention window sizes based on normalized traffic demands, the method significantly improves aggregate throughput while maintaining fairness comparable to classical PF, outperforming both standard DCF and conventional PF in simulations.

ABSTRACT

This paper presents a modified proportional fairness (PF) criterion suitable for mitigating the extit{rate anomaly} problem of multirate IEEE 802.11 Wireless LANs employing the mandatory Distributed Coordination Function (DCF) option. Compared to the widely adopted assumption of saturated network, the proposed criterion can be applied to general networks whereby the contending stations are characterized by specific packet arrival rates, $λ_s$, and transmission rates $R_d^{s}$. The throughput allocation resulting from the proposed algorithm is able to greatly increase the aggregate throughput of the DCF while ensuring fairness levels among the stations of the same order of the ones available with the classical PF criterion. Put simply, each station is allocated a throughput that depends on a suitable normalization of its packet rate, which, to some extent, measures the frequency by which the station tries to gain access to the channel. Simulation results are presented for some sample scenarios, confirming the effectiveness of the proposed criterion.

Motivation & Objective

• To address the rate anomaly problem in multirate IEEE 802.11 DCF under non-saturated traffic conditions, where classical fairness criteria fail due to unequal station loads.
• To develop a throughput allocation mechanism that incorporates actual station packet arrival rates ($\lambda_s$) and transmission rates ($R_d^s$), moving beyond the saturated network assumption.
• To enhance aggregate network throughput while preserving fairness levels comparable to classical proportional fairness.
• To optimize contention window sizes ($W_0^{(s)}$) per station based on traffic demand, ensuring efficient channel access without degrading fairness.

Proposed method

• Extends a bi-dimensional Markov model to account for multiple traffic classes based on channel occupancy duration, grouping stations with similar transmission durations into $N_c$ classes.
• Defines station-specific access probabilities ($\tau_s$) based on individual packet rates ($\lambda_s$) and transmission rates ($R_d^s$), enabling dynamic throughput allocation.
• Introduces a modified proportional fairness criterion that normalizes throughput allocation by station traffic demand, ensuring fairness relative to actual transmission frequency.
• Optimizes contention window sizes ($W_0^{(s)}$) per station using a linear programming framework (LPF) and a modified version (MLPF) that includes real traffic constraints.
• Uses a duration-class-based abstraction to simplify throughput analysis, assuming $N_c \leq N$ classes of channel occupancy durations.
• Employs simulation-based evaluation across multiple network scenarios to validate throughput and fairness improvements.

Experimental results

Research questions

• RQ1 How does the performance anomaly in multirate IEEE 802.11 DCF manifest under non-saturated traffic with heterogeneous packet arrival rates?
• RQ2 Can a modified proportional fairness criterion improve aggregate throughput without sacrificing fairness in general load conditions?
• RQ3 To what extent does incorporating actual station packet rates ($\lambda_s$) into the fairness metric enhance network performance compared to classical PF?
• RQ4 How does the proposed method compare to standard DCF and classical PF in terms of aggregate throughput and Jain’s fairness index?

Key findings

• The proposed MLPF (modified LPF) criterion achieves the highest aggregate throughput, reaching 4.69 Mbps in scenario A and 4.72 Mbps in scenario B, significantly outperforming standard DCF (1.89–1.93 Mbps).
• Jain’s fairness index improves to 0.9317 under MLPF in scenario A and 0.9290 in scenario B, approaching the fairness levels of classical PF while achieving higher throughput.
• The rate anomaly problem is greatly mitigated: in scenario B, where the two fastest stations have higher packet rates than the slowest, MLPF maintains higher throughput for faster stations compared to DCF.
• When the slowest station’s packet rate increases from 10 to 3300 pkt/s, DCF throughput drops sharply for faster stations, but MLPF maintains significantly higher aggregate throughput across all rates.
• The LPF and MLPF frameworks improve aggregate throughput over both standard DCF and classical PF, with MLPF showing the best trade-off between fairness and efficiency.
• The simulation results confirm that accounting for actual station traffic demand in the optimization framework leads to better throughput allocation, especially when high-rate stations are under heavy load.