The Ultimate Share-Everything PDES System

Abstract

The share-everything PDES (Parallel Discrete Event Simulation) paradigm is based on fully sharing the possibility to process any individual event across concurrent threads, rather than binding Logical Processes (LPs) and their events to threads. It allows concentrating, at any time, the computing power-the CPU-cores on board of a shared-memory machine-towards the unprocessed events that stand closest to the current commit horizon of the simulation run. This fruitfully biases the delivery of the computing power towards the hot portion of the model execution trajectory. In this article we present an innovative share-everything PDES system that provides (1) fully non-blocking coordination of the threads when accessing shared data structures and (2) fully speculative processing capabilities-Time Warp style processing-of the events. As we show via an experimental study, our proposal can cope with hard workloads where both classical Time Warp systems-based on LPs to threads binding-and previous share-everything proposalsnot able to exploit fully speculative processing of the events-tend to fail in delivering adequate performance.

Optimizing Simulation on Shared-Memory Platforms: The Smart Cities Case

“…The generality of this paradigm is also such that restoring a previous non-consistent state can be supported by different methodologies, e.g. traditional state saving and restore (Ianni et al 2018), reverse scrubbing (Cingolani et al 2016), or transactional memory (Santini et al 2015).…”

“…For a thorough discussion of the architectural details of a reference share-everything simulation platform, we refer the reader to the work in Ianni et al (2018). We discuss here the most fundamental aspects of the Ianni, Marotta, Cingolani, Pellegrini, and Quaglia reference implementation, for the sake of completeness.…”

“…Indeed, the share-everything paradigm has been devised with small granularity events in mind. Nevertheless, the effectiveness of this paradigm in previous applications shown in the literature, such as in (Ianni et al Ianni, Marotta, Cingolani, Pellegrini, and Quaglia 2018) suggested that it can be effectively exploited also in scenarios which a much coarser grain event granularity.…”

“…All this body of work has given rise to the share-everything paradigm, which has been recently specified in Ianni et al (2018). This work has shown how it is possible, on a shared-memory machine, to significantly reduce the number of events that are undone in the speculative simulation portion.…”

Hardware-Assisted Incremental Checkpointing in Speculative Parallel Discrete Event Simulation

“…Speculative Parallel Discrete event Simulation (PDES) is known to be a core method for deliverying high performance of model execution (Jefferson 1985), and for enabling the full exploitation of the available computing resources in both distributed and shared memory settings (Barnes et al 2013;Ianni et al 2018). At the same time, a core aspect to be considered when building speculative PDES platforms is the ability to reconstruct past simulation states whenever a causality violation-caused by wrong speculation paths-needs to be undone.…”