[Paper Review] Delay Analysis of Hybrid WiFi-LiFi System

Sihua Shao, Abdallah Khreishah+1 more|arXiv (Cornell University)|Oct 2, 2015

Optical Wireless Communication Technologies5 references3 citations

TL;DR

This paper analyzes delay performance in hybrid WiFi-LiFi systems, comparing non-aggregated (request routed to one of two networks) and aggregated (request split across both) configurations. Under Poisson arrivals and exponential request sizes, it proves analytically that bandwidth aggregation reduces minimum average system delay, and proposes an efficient approximation for the optimal splitting ratio with less than 3% delay penalty compared to the optimal solution.

ABSTRACT

Heterogeneous wireless networks are capable of effectively leveraging different access technologies to provide a wide variety of coverage areas. In this paper, the coexistence of WiFi and visible light communication (VLC) is investigated as a paradigm. The delay of two configurations of such heterogeneous system has been evaluated. In the first configuration, the non-aggregated system, any request is either allocated to WiFi or VLC. While in the second configuration, the aggregated system, each request is split into two pieces, one is forwarded to WiFi and the other is forwarded to VLC. Under the assumptions of Poisson arrival process of requests and the exponential distribution of requests size, it is mathematically proved that the aggregated system provides lower minimum average system delay than that of the non-aggregated system. For the non-aggregated system, the optimal traffic allocation ratio is derived. For the aggregated system, an efficient solution for the splitting ratio is proposed. Empirical results show that the solution proposed here incurs a delay penalty (less than 3\%) over the optimal result.

Motivation & Objective

• To evaluate system delay in hybrid WiFi-LiFi networks under varying configurations.
• To compare non-aggregated and aggregated system architectures in terms of average system delay.
• To derive the optimal traffic allocation ratio for the non-aggregated system.
• To propose an efficient approximation for the optimal request splitting ratio in the aggregated system.
• To empirically validate the delay benefits of bandwidth aggregation over non-aggregation.

Proposed method

• Models the system using M/D/1 queues for both WiFi and VLC links under Poisson request arrivals and exponentially distributed request sizes.
• Derives the average system delay for the non-aggregated system as a function of traffic allocation ratio α.
• Formulates the average system delay for the aggregated system as E[max(D_WiFi, D_VLC)], using order statistics to express E[min(D_WiFi, D_VLC)].
• Proposes an approximation method to estimate the optimal splitting ratio α in the aggregated system, minimizing delay under practical constraints.
• Employs simulation and analytical modeling to compare the approximated aggregated system with the non-aggregated system across varying parameters (λ, μ, B₁, B₂, N).
• Uses the ratio of approximated minimum delay in the aggregated system to the minimum delay in the non-aggregated system as a performance metric.

Experimental results

Research questions

• RQ1 Does bandwidth aggregation in a hybrid WiFi-LiFi system reduce the minimum average system delay compared to non-aggregated routing?
• RQ2 What is the optimal traffic allocation ratio in the non-aggregated system that minimizes average system delay?
• RQ3 Can an efficient approximation method be developed for the optimal request splitting ratio in the aggregated system with minimal delay penalty?
• RQ4 How do system parameters such as traffic load (λ), request processing rate (μ), and bandwidth (B₁, B₂) affect the delay performance of aggregated and non-aggregated systems?
• RQ5 What is the impact of the number of VLC access points (N) on the delay reduction gain from aggregation?

Key findings

• The aggregated system achieves a lower minimum average system delay than the non-aggregated system, as proven analytically using order statistics.
• The proposed approximation for the optimal splitting ratio in the aggregated system incurs a delay penalty of less than 3% compared to the theoretical optimum.
• The maximum additional delay due to the approximation is 2.7% when varying the number of VLC APs (N) from 1 to 10.
• The delay reduction from aggregation is most significant when WiFi bandwidth is small and VLC bandwidth is large, with gains diminishing as N increases or as λ and μ rise.
• The approximated minimum average system delay of the aggregated system is at least 16% lower than that of the non-aggregated system under the tested simulation parameters.
• The delay penalty is minimized when λ ≈ 0.33, μ ≈ 58, and B₁ ≈ 70, indicating optimal operating points for the approximation.