On Temporal Composability of Cyber-Physical Disruption-Tolerant Networks

Disruption-Tolerant Networks (DTNs) enable transfer of data when network nodes are only intermittently connected. With respect to the scarcity of connectivity in DTNs, a key research challenge is to guarantee schedulability of data ???ows which are to be fed into the network and prevent network resources from being channeled into serving data packets which cannot make deadline. Research efforts in real-time community are all concerned with traditional type of resources which are not available in time, but available in space. However, the nature of DTNs resources deviates from this model: not only are the network resources temporally unavailable, but also spatially unavailable. This distinct paradigm of resources demands for specialized schedulability analysis techniques to be developed. To this end, among a wide range of problems real-time DTNs require to be studied, in this paper we take the ???rst step in this direction and analyze schedulability of the workload in recurrent DTNs. We extend the recently proposed delay composition framework for disruption-tolerant networks and provide a worst-case estimation of end-to-end latency of data ???ows and investigate how latency is composed in DTNs. We then analyze whether or not the multi-criticality spatially-distributed data ???ows meet their end-to-end deadline constraints. Our study through simulation based experiments show that our approach has moderate pessimism and provides appropriate schedulability guarantees prior to establishment of the distributed workload.unpublishednot peer reviewe