Abstract. Overlay networks have emerged as a powerful paradigm to realise a large range of distributed services. However, as the number of overlays grows and the systems that use them become more intercon-nected, overlays must increasingly co-exist within the same infrastruc-ture. When this happens, overlays have to compete for limited resources, which causes negative interferences. This paper takes an opposite view, and argues that coexisting overlays may also introduce positive syner-gies that can be exploited to benefit a distributed system. Unfortunately, and in spite of some pioneering work, this phenomenon is still poorly un-derstood and has yet to be investigated systematically. To address this problem, this paper proposes a principled classification of synergies, and illustrates how it can be used to exploit synergies in a typical overlay platform targeting gossip protocols (GossipKit). We review in detail the risks and benefits of each identified synergy; we present experimen-tal data that validate their added value, and finally discuss the lessons we have learnt from our implementation. Key words: coexistence, synergy, gossip, overlay framework 1

Abstract—Managing wireless sensor networks in an energyefficient manner is no mean feat. Management requests and responses create additional traffic in addition to the data issuing from the network’s actual sensing application. Effective management therefore requires balancing the need for detailed oversight of the network against the energy consumption of the management system itself. This paper explores whether sending the management data and the sensing data together rather than separately can reduce the management system’s energy footprint. From the results of our experiment using BMAC and DYMO on MICAz motes running on TinyOS, we find that our approach does indeed substantially reduce the communication costs of the management system. We discuss different models for cooperation between the management system and the sensing application and estimate the potential trade-off between the number of packet transmissions and the delay of management data. To put the theoretical results into practice, we present a management framework for monitoring wireless sensor networks that is independent of the actual sensing application. This framework allows fine-grained control over the latency tolerated for management requests for the sake of reduced energy consumption. Measurements based on a prototype implementation of the framework in an experimental setup show that up to 61 % of the energy previously needed for management requests can be saved. I.