[Paper Review] Modeling Network Coded TCP: Analysis of Throughput and Energy Cost

Minji Kim, T.E. Klein+3 more|arXiv (Cornell University)|Aug 15, 2012

Network Traffic and Congestion Control4 references4 citations

TL;DR

This paper proposes TCP/NC, a network-coded variant of TCP that improves throughput and energy efficiency in lossy wireless networks by masking random erasures, enabling faster window growth and sustained high throughput even under packet loss. The key result shows TCP/NC reduces the number of required base stations by up to 50% compared to standard TCP, improving spectral efficiency and lowering operational costs without sacrificing user quality of service.

ABSTRACT

We analyze the performance of TCP and TCP with network coding (TCP/NC) in lossy networks. We build upon the framework introduced by Padhye et al. and characterize the throughput behavior of classical TCP and TCP/NC as a function of erasure probability, round-trip time, maximum window size, and duration of the connection. Our analytical results show that network coding masks random erasures from TCP, thus preventing TCP's performance degradation in lossy networks. It is further seen that TCP/NC has significant throughput gains over TCP. In addition, we show that TCP/NC may lead to cost reduction for wireless network providers while maintaining a certain quality of service to their users. We measure the cost in terms of number of base stations, which is highly correlated to the energy, capital, and operational costs of a network provider. We show that increasing the available bandwidth may not necessarily lead to increase in throughput, particularly in lossy networks in which TCP does not perform well. We show that using protocols such as TCP/NC, which are more resilient to erasures, may lead to a throughput commensurate the bandwidth dedicated to each user.

Motivation & Objective

• To analyze the performance of standard TCP and TCP/NC in lossy wireless networks with respect to throughput and energy efficiency.
• To investigate how network coding mitigates TCP's performance degradation caused by random erasures in wireless environments.
• To quantify the impact of erasure probability, RTT, window size, and connection duration on throughput for both TCP and TCP/NC.
• To model the number of base stations (N_bs) required to support a given user throughput, linking protocol choice to network cost and resource utilization.
• To demonstrate that TCP/NC can achieve higher effective throughput per unit bandwidth, reducing infrastructure and energy costs for providers.

Proposed method

• Adopts and extends the TCP model from Padhye et al. to analyze steady-state throughput under erasure conditions.
• Develops analytical models for TCP and TCP/NC throughput as functions of erasure probability (p), RTT, maximum window size (W_max), and connection duration.
• Uses NS-2 simulations to validate the analytical throughput predictions for both TCP and TCP/NC across varying network conditions.
• Introduces a cost model based on the number of base stations (N_bs) required to deliver a target user throughput, with N_bs proportional to energy, capital, and operational costs.
• Derives expressions for N_bs as a function of bandwidth (B), user throughput (T), and erasure probability (p), showing that TCP/NC maintains high T even at high p.
• Compares TCP and TCP/NC performance across multiple RTT and bandwidth configurations to assess robustness and scalability.

Experimental results

Research questions

• RQ1 How does network coding in TCP/NC improve throughput compared to standard TCP in lossy wireless networks?
• RQ2 To what extent does TCP/NC maintain high window size and throughput under increasing erasure probabilities?
• RQ3 What is the relationship between allocated bandwidth and actual user throughput for TCP and TCP/NC, especially under high loss rates?
• RQ4 How does the number of required base stations (N_bs) scale with erasure probability and protocol choice, and can TCP/NC reduce infrastructure costs?
• RQ5 Can TCP/NC achieve throughput close to the theoretical bandwidth limit (B ≈ T) even under high packet loss, and what does this imply for network resource efficiency?

Key findings

• TCP/NC achieves significantly higher steady-state throughput than standard TCP, especially under high erasure probabilities (e.g., p = 0.05), due to network coding’s ability to mask random losses.
• At p = 0.05, TCP/NC maintains throughput close to the theoretical bandwidth limit (B ≈ T), while standard TCP’s throughput drops sharply, leading to a 50% increase in required base stations (N_bs).
• For RTT = 100 ms and W_max = 50, the maximum achievable throughput is approximately 0.8 Mbps when RTT increases to 500 ms, limiting gains even with higher bandwidth.
• When RTT is 100 ms and W_max = 50, TCP/NC maintains a stable N_bs value regardless of erasure probability, reflecting robustness and consistent performance.
• The number of base stations required (N_bs) for TCP/NC remains close to the ideal case (B = T) even at high p, indicating efficient use of network resources.
• TCP/NC reduces the number of base stations needed to support a given user throughput by up to 50% compared to standard TCP, especially under high loss conditions, leading to lower energy and capital costs.