Repeatability in real-time distributed simulation executions

Middleware for real‐time distributed simulations

“…This is due to the likelihood that in some cases the messages do not arrive in time to be processed in one cycle of tick() calls. This behavior was also noted in [3] and is common for time-stepped execution models.…”

“…The first main issue is the completion of an LBTS computation by a specified deadline. Our real-time timemanagement implementation builds upon previous work in achieving repeatable real-time distributed simulations [3]. A primary challenge in repeatable performance is the timeliness is the tick() function.…”

On reproducibility and traceability of simulations

“…In the introduction we mentioned several works on protocols to solve the non-determinism issues of parallel execution (Fujimoto 2000) or real-time interactions (McLean and Fujimoto 2000). If we except these determinismrelated issues, repeatability is mainly a question of ensuring the long-term execution of a program.…”

Simulation for testing software agents - an exploration based on James

“…(Bacon and Goldstein 1991) classified non-determinism arising from input-data, system calls, and interrupts. Non-determinism arising from input data can be distinguished whether the input has been sent by logical processes, by externally running software, or by other sources, like human operators (McLean and Fujimoto 2000). Whereas in the context of Mole and for testing different planning agents (Uhrmacher and Gugler 2000) parallel, distributed simulators have been employed, for testing Autominder in the current virtual environment a sequential simulation suffices.…”

“…Currently we are designing a paced distributed simulation system which uses time-stamped messages. The underlying concept is similar to (McLean and Fujimoto 2000) in utilizing time stamps for reducing non-determinism.…”