Impact of Proposed Research

The work proposed here will have wide impact to a variety of computing fields. On the applications side, these network models will allow networking researchers the ability to radically re-think the set of experiments they need to perform in order to understand a particular network characteristic. The real power of PRC is not the speed of an individual simulation run but its ability to execute models forwards and backwards. As discussed in the Design of Experiments research thrust, this capability allows the modeler to ask questions about the model that can be answered in a few simple runs and not necessarily require an exhaustive $2^K$ factor simulation study. It this modeling power that we believe makes PRC worth the potential complexities and could be a huge payoff.

As previous indicated, not all computations are perfectly reversible. Many computations will generate entropy at the individual code statement or assembly language level. However, as we observed in Property 3 of Section 3, that just because a sequence of instructions are not reversible (as in the case of random number generators) does not mean that the higher-level functionality or model is not. It is this point we want to emphasize and is a fundamental research question - i.e., finding those higher-level reversible model representations. We believe because of Property 3, PRC will have a much wider applicability than one would imagine at first glance.

Moreover, if we are successful with network models, we believe this technique will be of value to any domain where discrete-event simulation is employed as analysis/decision support tool. We admit the topics proposed here are high risk, but the pay offs presented here are very much worth it.

For the parallel simulation community, this work represents a paradigm shift in the modeling and simulation of network systems. As previously indicated most of the research in network models use a conservative synchronization approach and optimistic approaches among those researchers are not seen as being viable. This research project will change that prevailing opinion.

In the area of quantum computing, this research will provide a means to explore the development of simulation models that could later be realized on a quantum computer.

In the area of general purpose parallel processing, this research will explore how to take general applications that have been transformed and made reversible and then attempt to automatically parallelize them using a Virtual Time Machine synchronization approach.