[Paper Review] Buffer Occupancy and Delivery Reliability Tradeoffs for Epidemic Routing

Pin‐Yu Chen, Meng-Hsuan Sung+1 more|arXiv (Cornell University)|Jan 24, 2016

Opportunistic and Delay-Tolerant Networks30 references3 citations

TL;DR

This paper analyzes the tradeoff between buffer occupancy and delivery reliability in epidemic routing for intermittently connected networks using a susceptible-infected-recovered (SIR) model. It proposes optimal global timeout and antipacket dissemination schemes, showing that per-node buffer occupancy depends only on packet loss rate and pairwise meeting rate under optimal timeout, and that cooperative antipacket dissemination drastically reduces buffer usage while ensuring reliable delivery.

ABSTRACT

To achieve end-to-end delivery in intermittently connected networks, epidemic routing is proposed for data delivery at the price of excessive buffer occupancy due to its store-and-forward nature. The ultimate goal of epidemic routing protocol design is to reduce system resource usage (e.g., buffer occupancy) while simultaneously providing data delivery with statistical guarantee. Therefore the tradeoffs between buffer occupancy and data delivery reliability are of utmost importance. In this paper we investigate the tradeoffs for two representative schemes: the global timeout scheme and the antipacket dissemination scheme that are proposed for lossy and lossless data delivery, respectively. For lossy data delivery, we show that with the suggested global timeout value, the per-node buffer occupancy only depends on the maximum tolerable packet loss rate and pairwise meeting rate. For lossless data delivery, we show that the buffer occupancy can be significantly reduced via fully antipacket dissemination. The developed tools therefore offer new insights for epidemic routing protocol designs and performance evaluations.

Motivation & Objective

• To address the critical tradeoff between high buffer occupancy and reliable data delivery in epidemic routing for delay-tolerant networks (DTNs).
• To develop analytical models that quantify buffer occupancy and delivery reliability under different immunity mechanisms in epidemic routing.
• To provide scalable, closed-form solutions for optimal global timeout values that guarantee a maximum tolerable packet loss rate.
• To evaluate the effectiveness of antipacket dissemination in minimizing buffer occupancy for lossless data delivery.
• To offer design guidelines for epidemic routing protocols that balance resource usage and delivery reliability.

Proposed method

• Models epidemic routing dynamics using the susceptible-infected-recovered (SIR) epidemiological framework.
• Derives state equations for the evolution of infected (data-carrying) and recovered (data-deleted) nodes under the global timeout scheme.
• Develops a closed-form expression for the optimal global timeout value that ensures the maximum tolerable packet loss rate is not exceeded.
• Analyzes the antipacket dissemination scheme where nodes proactively share 'antipackets' to trigger early deletion of redundant data.
• Uses ordinary differential equations (ODEs) to capture macroscopic spatiotemporal data dissemination and buffer dynamics.
• Validates models through simulations to confirm accuracy in predicting delivery and buffer behavior.

Experimental results

Research questions

• RQ1 What is the optimal global timeout value that guarantees a specified maximum packet loss rate in epidemic routing?
• RQ2 How does per-node buffer occupancy scale with network size under the global timeout scheme?
• RQ3 To what extent can cooperative antipacket dissemination reduce buffer occupancy in lossless epidemic routing?
• RQ4 How do the pairwise meeting rate and packet loss rate jointly influence buffer occupancy in epidemic routing?
• RQ5 Can the proposed models accurately predict system-level data delivery and buffer dynamics in DTNs?

Key findings

• The optimal global timeout value ensures that per-node buffer occupancy depends only on the maximum tolerable packet loss rate and pairwise meeting rate, and is independent of the total number of nodes, proving scalability.
• With the proposed global timeout, the system achieves reliable delivery while minimizing buffer usage, making it suitable for large-scale DTNs.
• The antipacket dissemination scheme significantly reduces buffer occupancy by enabling early deletion of redundant packets through cooperative immunity.
• Simulation results confirm that the analytical models accurately predict both data delivery dynamics and buffer occupancy trends in intermittently connected networks.
• The global timeout scheme is scalable and robust, as buffer occupancy remains bounded regardless of network size when the optimal timeout is applied.
• The antipacket dissemination mechanism enables lossless delivery with substantially lower buffer consumption compared to standard epidemic routing.