#discrete-event #events #distributed #conservative #numbers #framework

dvcompute_cons

Discrete event simulation library (conservative distributed simulation)

4 stable releases

2.0.0 Jun 29, 2024
1.3.5 Jan 12, 2022
1.3.4 Jan 6, 2022
1.3.3 Jan 5, 2022

#54 in Simulation


Used in 3 crates

MPL-2.0 license

625KB
15K SLoC

dvcompute_cons

This crate is a part of discrete event simulation framework DVCompute Simulator (registration number 2021660590 of Rospatent). The dvcompute_cons crate is destined for conservative distributed simulation, but the same code base is shared by the dvcompute crate destined for sequential simulation.

There are the following main crates: dvcompute (sequential simulation), dvcompute_dist (optimistic distributed simulation), dvcompute_cons (conservative distributed simulation) and dvcompute_branch (nested simulation). All four crates are very close. They are based on the same method.

Requirements

In case of conservative distributed simulation, it is expected that the dvcompute_mpi and dvcompute_core_cons dynamic (shared) libraries can be found by the operating system, when launching the executable file of the simulation model.

You can build the dvcompute_mpi library yourself from the sources that require CMake, C++ compiler and some MPI implementation that you are going to use.

But the dvcompute_core_cons dynamic library must satisfy the predefined binary interface (it must implement the event queue). You can request the author for the pebuilt version that implements this interface. This prebuilt version is a part of "Redistributable Code" portions of DVCompute Simulator.

Simulation Method

The main idea is to use continuations for modeling discontinuous processes. Such continuations are themselves wrapped in the monad, for which there are easy-to-use combinators. This idea is inspired by two sources: (1) combinators for futures that were in Rust before introducing the async/await syntax and (2) the Aivika simulation library that I developed in Haskell before.

Here is an example that defines a model of the machine that breaks down and then it is repaired:

const UP_TIME_MEAN: f64 = 1.0;
const REPAIR_TIME_MEAN: f64 = 0.5;

fn machine_process(total_up_time: Grc<RefComp<f64>>) -> ProcessBox<()> {
    let total_up_time2 = total_up_time.clone();
    random_exponential_process(UP_TIME_MEAN)
        .and_then(move |up_time| {
            RefComp::modify(total_up_time, move |total_up_time| {
                total_up_time + up_time
            })
            .into_process()
            .and_then(move |()| {
                random_exponential_process_(REPAIR_TIME_MEAN)
            })
            .and_then(move |()| {
                machine_process(total_up_time2)
            })
        })
        .into_boxed()
}

These computations are combined with help of the monadic bind. Such computations should be run later to take effect.

Examples

You can find examples in the author's repository.

Documentation

Tutorial

Also you can read the PDF document DVCompute Simulator Tutorial.

Bibliography

  • Sorokin David. DVCompute Simulator for discrete event simulation. Prikladnaya informatika=Journal of Applied Informatics, 2021, vol.16, no.3, pp.93-108 (in Russian). DOI: 10.37791/2687-0649-2021-16-3-93-108

Licence

Copyright 2020-2024 David Sorokin davsor@mail.ru, based in Yoshkar-Ola, Russia

This software is subject to the terms of the Mozilla Public License, v. 2.0. If a copy of the MPL was not distributed with this file, You can obtain one at http://mozilla.org/MPL/2.0/.

Dependencies

~1.8–2.7MB
~51K SLoC