This paper reviews issues concerning the design of adaptive protocols for parallel discrete event simulation (PDES). The need for adaptive protocols are motivated in the background of the classical synchronization problem that has driven much of the research in this field. Traditional conservative and optimistic protocols and their hybrid variants --- that form the basis of adaptive protocols --- are also discussed. Adaptive synchronization protocols are reviewed with special reference to their characteristics regarding the aspects of the simulation state that influence the adaptive decisions and the control parameters used. Finally, adaptive load management strategies and their relationship to the synchronization protocol are discussed. Keywords: Simulation, Computers, Methodology.

In message passing environments, the message send time is dominated by overheads that are relatively independent of the message size. Therefore, fine-grained applications (such as Time-Warp simulators) suffer high overheads because of frequent communication. In this paper, we investigate the optimization of the communication subsystem of Time-Warp simulators using dynamic message aggregation. Under this scheme, Time-Warp messages with the same destination LP, occuring in close temporal proximity are dynamically aggregated and sent as a single physical message. Several aggregation strategies that attempt to minimize the communication overhead without harming the progress of the simulation (because of messages being delayed) are developed. The performance of the strategies is evaluated for a network of workstations, and an SMP, using a number of applications that have different communication behavior. 1 Introduction In distributed environments the performance of the communication subsy...

Distributedsynchronizationforparallelsimulationisgenerallyclassifiedasbeingeitheroptimisticorconservative. Whileconsiderableinvestigationshavebeenconducted toanalyzeandoptimizeeachofthesesynchronization strategies,verylittlestudyonthedefinitionandstrictness ofcausalityhavebeenconducted.Dowereallyneed topreservecausalityinalltypesofsimulations?This paperattemptstoanswerthisquestion.Wearguethat significantperformancegainscanbemadebyreconsideringthisdefinitiontodecideiftheparallelsimulation needstopreservecausality.Weinvestigatethefeasibility ofunsynchronizedparallelsimulationthroughtheuseof severalqueuingmodelsimulationsandpresentacomparativeanalysisbetweenunsynchronizedandTimeWarp simulation.

ABSTRACT In this paper we describe an adaptive, optimistic synchronisation mechanism for the parallel discrete event simulation of agent-based systems. The mechanism uses the Sphere of Influence (SoI) of an event (the region of the shared simulation state read or written to by the event) to define an adaptive metric which can be used with a throttling mechanism such as moving time windows. We show how such a metric can be calculated by monitoring the common reads and writes made by the agents to the shared simulation state modelling the agent’s environment, and present the results of our preliminary investigations into the relationships between agent read and write patterns and rollback frequency.

It is well known that Time Warp may suffer from poor performance due to excessive rollbacks caused by overly optimistic execution. Here we present a simple flow control mechanism using only local information and GVT that limits the number of uncommitted messages generated by a processor, thus throttling overly optimistic TW execution. The flow control scheme is analogous to traditional networking flow control mechanisms. A "window" of messages defines the maximumnumber of uncommitted messages allowed to be scheduled by a process. Committing messages is analogous to acknowledgments in networking flow control. The initial size of the window is calculated using a simple analytical model that estimates the instantaneous number of messages that a process will eventually commit. This window is expanded so that the process may progress up to the next commit point (generally the next fossil collection), and to accommodate optimistic execution. The expansions to the window are based on monitori...

This paper presents a sparse state saving scheme for Time Warp parallel discrete event simulation. The scheme bases the selection of the states to be recorded on the event history of the logical processes. To this purpose, statistics on the virtual time advancement of the processes are collected for the prediction of virtual time intervals that are likely to contain rollback points; the states corresponding to the starting point of those intervals are recorded as checkpoints in order to reduce the average coasting forward. The percentage of states to be recorded is defined by a parameter whose value is dynamically recalculated on the basis of the on-line observation of the variation of a checkpointing-rollback cost function. Simulation results of synthetic workloads are presented for a performance comparison with previous schemes. 1

This paper presents a checkpointing scheme for optimistic simulation which is a mixed approach between periodic and probabilistic checkpointing. The latter, basing on statistical data collected during the simulation, aims at recording as checkpoints states of a logical process that have high probability to be restored due to rollback (this is done in order to make those states immediately available) . The periodic part prevents performance degradation due to state reconstruction (coasting forward) cost whenever the collected statistics do not allow to identify states highly likely to be restored. Hence, this scheme can be seen as a highly general solution to tackle the checkpoint problem in optimistic simulation. A performance comparison with previous solutions is carried out through a simulation study of a store-and-forward communication network in a two-dimensional torus topology. 1 Introduction Parallel discrete event simulation aims at speeding up the simulation process of comple...

Abstract not found

Wepresentanoveladaptivesynchronizationalgorithm, calledtheminimumaveragecost(MAC)algorithm, inthecontextoftheParaSolparallelsimulationsystem. ParaSolisamultithreadedsystem forparallelsimulationonshared-anddistributedmemoryenvironments, designedtosupportdomainspecificSimulationObjectLibraries. Theproposed MACalgorithmisbasedonminimizingthecost ofsynchronizationdelayandrollbackataprocess, wheneveritssimulationdrivermustdecidewhether toeitherproceedoptimisticallyortodelayprocessing. Intheformercasetheriskisrollbackcost,inthe eventofastraggler'sarrival.Inthelattercasetherisk isunnecessarydelay,intheeventalate-comerisnot astraggler.InadditiontotheMACalgorithmand anoptimaldelaycomputationmodel,wereporton someearlyexperimentscomparingtheperformance ofMAC-basedadaptivesynchronizationtooptimistic synchronization. 1

Abstract not found