The use of parallel computers for space-network modeling
J. Hant, D. Lanzinger, M. Coodey · 2005
Cost-effective parallel computing such as Beowulf clusters offer a new technology that can be employed for computer event modeling (simulators or emulators). The Beowulf network evaluation tool (BeoNET) is a network-modeling tool designed to address the specific issues of space networks. It includes a user-end problem definition subsystem that runs on desktop PCs and a run-time subsystem that runs on a Beowulf cluster, consisting of possibly hundreds of processors connected via a high-speed, dedicated LAN. In its standard configuration, it uses a one-to-one mapping of modeled network nodes (i.e. satellites, ground stations) to processors, and a one-to-one mapping modeled network messages to BeoNET messages. By taking advantage of parallelism, it can achieve run-times that are faster than real time. By running the network emulation in software, the human time to reconfigure the emulation and analyze results is minimized, allowing the assessment of many possible solutions. Finally, by having a one to one mapping of modeled nodes and messages with BeoNET internal nodes and messages, BeoNET can achieve a level of fidelity that many network simulators, which rely on aggregation and abstraction, cannot.