Modern distributed simulations require a substantial network infrastructure – be it satellite links, broadband internet, radio communications, or a mix of all of these. Modern, large-scale live, virtual, constructive (LVC) simulations may require high numbers both of human-in-the-loop participants and constructed physical entities, each of which must communicate in some way with the others. Thus, when planning and executing a large-scale distributed LVC simulation it is extremely beneficial to have a good idea of whether your network is capable of facilitating the necessary data transfer and communications quickly enough to enable real-time reactions to perform the particular mission. We aim to provide a government-developed modeling and simulation tool which will allow predictions of an arbitrary network’s data transfer and communication capability, extensible to eventually include computational capability predictions. Currently, we are able to simulate an arbitrary number of applications and users interacting via simulation processes submitting jobs by broadcast and multicast methods through a specified network infrastructure to be analyzed on remote servers, with both RAM and SSD elements, jitter functionality, propagation delays, and network loading due to various real-world properties. Each link between network elements has a specified propagation delay (artificial latency), maximum transmission unit, and bandwidth. These elements are particularly important, as excessive latency can be greatly detrimental to the processing requirements of large-scale distributed simulation events. We believe that the behavior of the model as the network parameters are varied is reasonable, and that this model could be useful in predicting capabilities of larger distributed networks. We support this conclusion with data taken from a small test network architecture modelled after a development enclave in our lab.
Simulation
Predictive Performance Modeling for Distributed Live, Virtual, Constructive Environments
16 views