DTN and Satellite Communications

Satellite communications are an interesting and promising application field for Delay/Disruption Tolerant Networking (DTN). Although primarily conceived for deep space communications and sensor networks, DTN was immediately recognized as applicable to satellite environments, in particular to cope with the intermittent channels typical of LEO (Low Earth Orbit) constellation satellite systems. However, DTN applicability to the satellite field is not limited to LEOs, but extends to other systems, like GEOs (Geostationary Earth Orbit), where performance of reliable transport protocols, like TCP, is impaired by the characteristic of the satellite channel itself. Even in the most favorable case of continuous end-to-end connectivity, long delays inherent in GEO systems, together with the possible presence of a high Packet Error Rate (PER) due to the wireless channel, severely affect performance. The most common countermeasure is the insertion of intermediate agents, called PEPs (Performance Enhancing Proxies), which generally offer good performance but violate the end-to-end semantics of transport protocols. In this case too DTNs can have an important role. The possible applications of the DTN concept in satellite communications are various and can be conveniently classified on the basis of the presence or not of end-to-end connectivity: • continuous end-to-end connectivity; end-to-end connectivity is usually present; this may be the case with GEO satellite systems with terrestrial fixed terminals; • random intermittent end-to-end connectivity; end-to-end connectivity may be present but channel disruptions are frequent, and difficult, or impossible, to predict; for example, a GEO satellite connecting mobile terminals on means of transport, where the satellite link is disrupted by tunnels or other obstructions; • scheduled intermittent end-to-end connectivity; end-to-end connectivity is assured at regular intervals; for example, a LEO satellite for Earth observation, which can connect to its gateway stations only at intermittent but predictable intervals due to its orbital motion; • no end-to-end connectivity; there is never end-to-end connectivity between end-points; for example, a single LEO satellite acting as a “mule” between a terrestrial sensor network and a remote satellite gateway station, which are never in satellite visibility at the same time. These scenarios will be investigated in specific sections. However, let us briefly anticipate here the reasons why DTN can be more advantageous with respect to other approaches, like end-to-end transport protocols or PEPs. To this end, let us re-consider the four above scenarios in reverse order, as the last is the most suited to DTN. In the last scenario the absence of end-to-end connectivity prevents the establishment of TCP, or TCP-like, connections. Even unreliable UDP transfers are impossible, due the lack of a continuous path between end nodes, and the only possible approach is the application of DTN “store-and forward” techniques. Data must first be transferred on intermediate nodes, then, when possible, transferred to the receiver. This task can easily be accomplished by DTN, unlike PEPs or end-to-end transport protocols. The third scenario (intermittent end-to-end connectivity) is the next most suited, as end-to-end transfers are now possible, but only at scheduled times and for a limited amount of time. This latter restriction poses a strict limit on the total data that can be transferred at each availability interval. For example, files exceeding the maximum data constraint must be divided into multiple parts. In this case DTN is preferable. If the DTN concept is implemented through the “bundle layer”, long files can be transferred by segmenting them in multiple bundles of the right dimension or, alternatively, by inserting the whole file in a single bundle, by relying on pro-active bundle segmentation. In the second scenario there is end-to-end connectivity but..