Hybrid performance modelling of opportunistic networks

We demonstrate the modelling of opportunistic networks using the process algebra stochastic HYPE. Network traffic is modelled as continuous flows, contact between nodes in the network is modelled stochastically, and instantaneous decisions are modelled as discrete events. Our model describes a network of stationary video sensors with a mobile ferry which collects data from the sensors and delivers it to the base station. We consider different mobility models and different buffer sizes for the ferries. This case study illustrates the flexibility and expressive power of stochastic HYPE. We also discuss the software that enables us to describe stochastic HYPE models and simulate them.

💡 Research Summary

This paper presents a comprehensive performance modelling framework for opportunistic networks using the stochastic HYPE process algebra, a hybrid formalism that seamlessly integrates continuous flows, stochastic events, and instantaneous discrete actions. Opportunistic networks are characterised by intermittent, unpredictable contacts between nodes, making traditional packet‑oriented simulators ill‑suited for capturing the simultaneous presence of continuous data generation, random meeting times, and rapid control decisions. Stochastic HYPE addresses this gap by representing continuous traffic as differential equations over continuous variables, modelling node contacts as stochastic events with time‑varying rates, and encoding instantaneous decisions (such as buffer overflow handling) as discrete events.

The authors illustrate the approach with a concrete case study: a set of stationary video sensors that continuously generate high‑volume video streams, and a mobile data‑collector (the “ferry”) that periodically visits the sensors, downloads their data, and forwards it to a base station. Each sensor’s video stream is modelled as a continuous variable (x_i(t)) governed by a simple ODE reflecting the sensor’s production rate. The ferry‑sensor contacts are modelled as stochastic events (C_{ij}) whose occurrence rates (\lambda_{ij}(t)) depend on the ferry’s mobility model. Three mobility models are examined – random walk, Lévy walk, and a deterministic patrol route – each yielding a distinct functional form for (\lambda_{ij}(t)).

Buffer management on the ferry is captured by instantaneous events. When the buffer reaches its capacity, a “buffer‑overflow” event fires, triggering a predefined policy such as packet drop, priority reshuffling, or forced transmission to the base station. Conversely, when the ferry reaches the base station, a “delivery‑complete” event empties the buffer and records the end‑to‑end delay for the transferred data. All these components are composed using stochastic HYPE’s algebraic operators, producing a single hybrid model that can be automatically translated into a set of ODEs coupled with an event scheduler.

The simulation environment built around stochastic HYPE automatically performs this translation, integrates the ODEs, and processes stochastic and instantaneous events in a unified time line. The authors conduct a factorial experiment varying (i) the ferry’s mobility model, (ii) its buffer size (10 MB, 50 MB, 100 MB, and unlimited), and (iii) the ferry’s speed. For each of the 12 configurations they measure three key performance indicators: average end‑to‑end delay, data loss ratio, and average data collected per ferry visit.

Key findings include:

1. Higher ferry speeds increase contact frequency, reducing average delay, but when buffer size is small the benefit is offset by a sharp rise in data loss because the buffer overflows more often.
2. Lévy‑walk mobility, which concentrates visits on a subset of sensors, yields lower delays for those sensors but higher variance across the network, whereas random‑walk mobility provides more uniform service at the cost of slightly higher overall delay.
3. Buffers of 100 MB or larger effectively eliminate data loss (below 1 % in all scenarios) and cause the delay curves to plateau, indicating that beyond a certain capacity the system becomes contact‑limited rather than buffer‑limited. In other words, sufficient buffering can compensate for sparse contact opportunities.

Beyond the quantitative results, the paper discusses methodological insights. The hybrid nature of stochastic HYPE allows modelers to modify or extend the system by adding new stochastic events (e.g., energy‑draining failures) or new continuous flows (e.g., adaptive video compression) with minimal changes to the existing specification. The automatic translation to ODEs and the integrated event scheduler dramatically reduce simulation setup time compared with building separate discrete‑event and continuous‑flow simulators.

In conclusion, stochastic HYPE proves to be a powerful and flexible language for modelling the intricate dynamics of opportunistic networks, where continuous data production, random encounters, and instantaneous control actions coexist. The case study validates the expressiveness of the formalism and demonstrates its practical utility for exploring design trade‑offs such as mobility patterns and buffer provisioning. The authors suggest future work extending the framework to multi‑ferry scenarios, energy‑aware policies, and validation against real‑world trace data, thereby broadening the applicability of stochastic HYPE to a wider class of delay‑tolerant and disruption‑tolerant networking systems.