Prioritized Epidemic Routing for Opportunistic Networks

Ramanathan, Hansen, Basu, Rosales-Hain, Krishnan

manet routing networking protocol epidemic metrics aa qos

@article{ramanathan:mobisys-wmo2007,
  title={Prioritized Epidemic Routing for Opportunistic Networks},
  author={Ram Ramanathan and Richard Hansen and Prithwish Basu and
          Regina Rosales-Hain and Rajesh Krishnan},
  journal={MobiSys Workshop on Mobile Opportunistic Networking},
  year={2007},
  pages={62--66}
}

Epidemic forwarding has very successful delivery rates

• Particularly true for low resource loads, sparse conectivity
• But struggles under high resource loads, notably for memory and bandwidth
• PREP tries to address this by prioritizing bundles for delivery and storage

Bundles are assigned transmit and drop (inverse) priorities

• When a link is discovered, lower valued bundles are transmitted first
  • Transmit priority is based on (lower) costs to destination via next link and bundle time-to-expire
• When storage buffers are nearly filled, bundles with higher values are dropped first
  • Bundles under a certain hop (forwarding) count are not dropped
  • Beyond that, bundles with larger shortest path cost to their destination are dropped first

The drop policy has the effect of clustering bundles around their destinations, such that they are more likely to be delivered as links become available

• Link quality is measured by availability ratios, measured by tracking status as determined by HELLOs
• Link state is propagated in an epidemic fashion, syncing tables with neighbors

This approach trades bandwidth consumption and latency for higher delivery rates

• Problematic with most sensor, MANET nodes, where battery lifetime is critical
• Is that trade in efficiency and latency worth it? Depends a lot on application

How do you manage independent priorities? What if several independent systems come together, how do you ensure their priorities will mesh well in a global way?