Simulation, Orchestration and Logical Clocks

Abstract

A language in which discrete event simulations can be coded needs to support the features (1) to describe behavior of a single physical process, (2) to describe concurrent ctivities of multiple physical processes, including communication, synchronization and interruption, (3) to account for passage of time, and (4) to record system state at appropriate points and create statistical summaries. Orc, a recent language for orchestration of distributed services, combines these features so that complex simulations can be expressed very succinctly. This talk describes the relevant features of Orc for simulation and illustrates them using a number of realistic examples.  Additionally, we show that certain combinatorial problems, such as shortest paths in graphs and many problems in computational geometry, can be cast as simulation problems, and solved very simply in Orc.