Modeling Optimal Retransmission Timeout Interval for Bundle Protocol

TLDR

Delay/disruption tolerant networking (DTN) was designed for reliable data delivery over long propagation delays and frequent link disruptions, especially in deep‑space missions where the core bundle protocol (BP) is essential. The study models an optimal retransmission‑timeout interval to maximize BP goodput in deep‑space links with long delays and high loss. The authors derive a mathematical model that computes the RTO timer interval that optimizes BP goodput, normalizing it against the total data load to ensure complete file delivery. Experimental results confirm that the modeled RTO interval yields the highest normalized BP goodput and accurately predicts performance across all tested scenarios.

Abstract

Delay/disruption tolerant networking (DTN) was proposed as a networking architecture for reliable data delivery despite an extremely long propagation delay and frequent/lengthy link disruptions. The challenging problem of mission control and data delivery in deep-space explorations is a typical application scenario of the DTN technology. Reliable data delivery of DTN relies heavily on its core bundle protocol (BP). Performance evaluation and improvement of BP for deep-space communications are presently underway. The setting of the retransmission time-out (RTO) timer of BP is critical for reliable and highly efficient file transfer in a deep-space communication environment. In this paper, we present analytical modeling of an optimal RTO timer interval for the best goodput performance of BP in deep-space communications characterized by a very long signal propagation delay and lossy data links. A model is developed to compute the RTO timer interval that will result in the best goodput performance of BP normalized with respect to the total data load transmitted in order to achieve successful delivery of an entire file. Experimental validation using a PC-based testbed indicates that the RTO timer interval indicated by the model achieves the best normalized goodput performance of BP. The model predicts performance in all the experimented scenarios.