Resource-aware capacity evaluation for heterogeneous, disruption-tolerant networks

Estimating end-to-end capacity is challenging in disruption-tolerant networks (DTNs) because reliable and timely feedback is usually unavailable. This paper proposes a resource-aware framework for estimating capacity between pairs of nodes. The proposed framework builds on information gathered autonomously by nodes so that results emerge from actual network properties. Achievable capacity is formulated as a linear programming problem. The objective function attempts to maximize delivered size when input data are quantized into messages at source and then injected in a single burst. Problem constraints emerge from conditioning intermediary nodes to avoid resource exhaustion. While the optimal solution remains challenging for most practical settings, the paper discusses alternatives that are less complex and more suitable for real implementation. We cover two scenarios: i) when network mobility is periodic or known in advance, and ii) when mobility is random but its spatial distribution remains stable in time. We claim that the proposed framework can be used to avoid resource exhaustion and network congestion in heterogeneous environments where no information about the system is known in advance. The results have been validated by simulations under various settings (number of flows, store-carry-forward protocols, homogeneous and heterogeneous resource distribution).