Parallel discrete-event simulation (PDES) is a well-established family of methods to accelerate discrete-event simulations. However, the available algorithms vary substantially in the performance achievable for different simulation models, largely preventing generic solutions applicable by modellers without expert knowledge. For instance, in Time Warp, the processing elements execute events asynchronously and speculatively with high aggressiveness, leading to frequent and costly rollbacks if misspeculations occur often. In contrast, synchronous approaches such as the new Window Racer algorithm exhibit a more cautious form of speculation. In the present paper, we combine these two fundamentally different algorithms within a single runtime environment, allowing for a choice of the best algorithm for different model segments. We describe the architecture and the algorithmic considerations to support the efficient coexistence and interaction of the algorithms without violating the correctness of the simulation. Our experiments using a synthetic benchmark and an epidemics model show that the hybrid algorithm is less sensitive to its configuration and can deliver substantially higher performance in models with varying degrees of coupling among entities compared to each algorithm on its own.

Piccione, A., Andelfinger, P., Pellegrini, A. (2023). Hybrid Speculative Synchronisation for Parallel Discrete Event Simulation. In Proceedings of the 2023 ACM SIGSIM Conference on Principles of Advanced Discrete Simulation (pp.84-95). New York : Association for Computing Machinery [10.1145/3573900.3591124].

Hybrid Speculative Synchronisation for Parallel Discrete Event Simulation

Alessandro Pellegrini
2023-06-01

Abstract

Parallel discrete-event simulation (PDES) is a well-established family of methods to accelerate discrete-event simulations. However, the available algorithms vary substantially in the performance achievable for different simulation models, largely preventing generic solutions applicable by modellers without expert knowledge. For instance, in Time Warp, the processing elements execute events asynchronously and speculatively with high aggressiveness, leading to frequent and costly rollbacks if misspeculations occur often. In contrast, synchronous approaches such as the new Window Racer algorithm exhibit a more cautious form of speculation. In the present paper, we combine these two fundamentally different algorithms within a single runtime environment, allowing for a choice of the best algorithm for different model segments. We describe the architecture and the algorithmic considerations to support the efficient coexistence and interaction of the algorithms without violating the correctness of the simulation. Our experiments using a synthetic benchmark and an epidemics model show that the hybrid algorithm is less sensitive to its configuration and can deliver substantially higher performance in models with varying degrees of coupling among entities compared to each algorithm on its own.
2023 ACM SIGSIM Conference on Principles of Advanced Discrete Simulation
Rilevanza internazionale
contributo
giu-2023
Settore ING-INF/05 - SISTEMI DI ELABORAZIONE DELLE INFORMAZIONI
English
Intervento a convegno
Piccione, A., Andelfinger, P., Pellegrini, A. (2023). Hybrid Speculative Synchronisation for Parallel Discrete Event Simulation. In Proceedings of the 2023 ACM SIGSIM Conference on Principles of Advanced Discrete Simulation (pp.84-95). New York : Association for Computing Machinery [10.1145/3573900.3591124].
Piccione, A; Andelfinger, P; Pellegrini, A
File in questo prodotto:
File Dimensione Formato  
Pic23b.pdf

solo utenti autorizzati

Tipologia: Documento in Post-print
Licenza: Copyright dell'editore
Dimensione 733.69 kB
Formato Adobe PDF
733.69 kB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2108/323429
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? 1
social impact