Relaxing Synchronization Constraints in Distributed Agent-based Simulations

Rihawi, Omar; Secq, Yann; Mathieu, Philippe

doi:10.11113/jt.v63.1957

Cited by 3 publications

(2 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We considered the characteristics of the agent-based models in traffic simulation and analyzed the discrepancy between the sequential simulation and the relaxed parallel simulation. A study on the impact of relaxing the barrier synchronization in distributed agent-based simulations was presented in [24], however, asynchronous synchronization was not mentioned and it was not studied how to maintain the accuracy of a distributed simulation.…”

Section: Related Workmentioning

confidence: 99%

An Asynchronous Synchronization Strategy for Parallel Large-scale Agent-based Traffic Simulations

Cai

Aydt³

et al. 2015

Proceedings of the 3rd ACM SIGSIM Conference on Principles of Advanced Discrete Simulation

View full text Add to dashboard Cite

Large-scale agent-based traffic simulation is a promising tool to study the road traffic and help solving traffic problems, such as congestion and high emission in megacities. Such simulation requires high computational resource which triggers the need for parallel computing. The parallelization of agent-based traffic simulations is generally performed by decomposing the simulation space into spatial subregions. The agent models contained by each subregion are executed by Logical Processes (LPs). As the simulated system evolves over the simulation time in individual LPs, synchronization among LPs is required due to data dependencies. Existing work has used global barriers for synchronization which is a type of synchronous synchronization method. However, global barriers have very low efficiency due to the waiting of processes at barriers. High synchronization overhead is still one of the major performance issues in parallel large-scale agent-based traffic simulations. In this paper, we proposed a novel asynchronous conservative synchronization strategy named Mutual Appointment (MA) to address this issue. MA removes global barriers and allows LPs to communicate individually. Since the efficiency of conservative synchronization relies on the lookahead of the simulated system, a heuristic was developed to increase the lookahead in agentbased traffic simulations. It takes advantage of the intrinsic uncertainties in traffic simulations. MA together with the lookahead heuristic forms the Relaxed Mutual Appointment (RMA) strategy. Its efficiency was investigated in the parallel agent-based traffic simulator SEMSim Traffic using real world traffic data. Experiment results showed that the MA Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from permissions@acm.org. strategy improved the speed-up of the parallel simulation compared to the barrier method, and the RMA strategy further improved the MA strategy by reducing the number of synchronization messages significantly.

show abstract

Section: Related Workmentioning

confidence: 99%

An Asynchronous Synchronization Strategy for Parallel Large-scale Agent-based Traffic Simulations

Cai

Aydt³

et al. 2015

Proceedings of the 3rd ACM SIGSIM Conference on Principles of Advanced Discrete Simulation

View full text Add to dashboard Cite

show abstract

“…It is intended to organise synchronisation operations into a synchronous manner which is more suitable for the GPU architecture. Another study is presented in [97] which studies the effect of relaxing the synchronisation of distributed agentbased simulations. Synchronisation operations are simply skipped.…”

Section: Relaxation Of Synchronisation Utilising Uncertaintymentioning

confidence: 99%

Load-balancing and synchronisation for parallel agent-based road traffic simulation

Xu¹

View full text Add to dashboard Cite

Workload balance and synchronisation of logical processes (LPs) are two critical factors that influence the execution time and the maximum speed-up of a parallel agent-based road traffic simulation. The original contributions of this thesis are the new methods that make use of the characteristics of agent models to efficiently conduct load-balancing and synchronisation in parallel agent-based road traffic simulations. We firstly evaluated the existing works of partitioning and load-balancing, as well as synchronisation protocols for parallel agent-based traffic simulations. We identified that the existing partitioning algorithms do not fully minimise the communication overhead between LPs. As for simulation protocols, most of the existing traffic simulators apply global barriers to synchronise LPs. Some existing works use conservative protocols to allow LPs to progress asynchronously. However, the performance may not be better than that of using global barrier, since in general agent models in traffic simulation only allow small lookahead. We then developed a graph partitioning algorithm, aiming to reduce the synchronisation overhead of the simulation while maintaining workload balance. In addition to balancing workload and minimizing communication volume, it also aims to minimise the number of neighbouring partitions of each partition. Minimising the number of neighbouring partitions can reduce synchronisation messages among LPs during the simulation. Less synchronisation messages means less communication overhead to the simulation. To further reduce the synchronisation overhead, two methods of improving the lookahead were developed. The first method takes advantage of the intrinsic uncertainties vii LR MA protocol relaxed by lowering the resolution of agent models MA mutual appointment RCB recursive coordinate bisection SE MA protocol relaxed by skipping and estimation xviii

show abstract

Relaxing Synchronization in Parallel Agent-Based Road Traffic Simulation

Cai

Aydt³

et al. 2017

ACM Trans. Model. Comput. Simul.

View full text Add to dashboard Cite

Large-scale agent-based traffic simulation is computationally intensive. Parallel computing can help to speed up agent-based traffic simulation. Parallelization of agent-based traffic simulations is generally achieved by decomposing the road network into subregions. The agents in each subregion are executed by a Logical Process (LP). There are data dependencies between LPs which require synchronization of LPs. An asynchronous protocol allows LPs to progress and communicate asynchronously. LPs use lookahead to indicate the time to synchronize with other LPs. Larger lookahead means less frequent synchronization operations. High synchronization overhead is still a major performance issue of large-scale parallel agent-based traffic simulations. In this article, two methods to increase the lookahead of LPs for an asynchronous protocol are developed. They take advantage of uncertainties in traffic simulation to relax synchronization without altering simulation results statistically. Efficiency of the proposed methods is investigated in the parallel agent-based traffic simulator SEMSim Traffic. Experiment results showed that the proposed methods are able to reduce overall running time of the parallel simulation compared to existing methods.

show abstract

Relaxing Synchronization Constraints in Distributed Agent-based Simulations

Cited by 3 publications

References 14 publications

An Asynchronous Synchronization Strategy for Parallel Large-scale Agent-based Traffic Simulations

An Asynchronous Synchronization Strategy for Parallel Large-scale Agent-based Traffic Simulations

Load-balancing and synchronisation for parallel agent-based road traffic simulation

Relaxing Synchronization in Parallel Agent-Based Road Traffic Simulation

Contact Info

Product

Resources

About