## Time-Parallel Simulation with Approximative State Matching (2004)

Venue: | In Proceedings of the 18th Workshop on Parallel and Distributed Simulation |

Citations: | 7 - 5 self |

### BibTeX

@INPROCEEDINGS{Kiesling04time-parallelsimulation,

author = {Tobias Kiesling and Siegfried Pohl},

title = {Time-Parallel Simulation with Approximative State Matching},

booktitle = {In Proceedings of the 18th Workshop on Parallel and Distributed Simulation},

year = {2004},

pages = {195--202}

}

### OpenURL

### Abstract

Time-Parallel Simulation offers the potential of massive parallelization of a simulation application, due to the amount of achievable parallelism not being restricted by the decomposability of the state space of a simulation model. Unfortunately, the potential speedup of a time-parallel simulation highly depends on the ability to match final and initial states of adjacent partitions. However, depending on the properties of the underlying simulation model, it might be feasible to accept a simulation iteration, even if the states of adjacent partitions do not match exactly. This leads to the concept of approximative state matching in time-parallel simulation, which is introduced in this paper. Experiments with a prototypical implementation of a simple simulation model show encouraging results in terms of simulation speedup and introduced error. 1

### Citations

708 |
Parallel discrete event simulation
- Fujimoto
- 1990
(Show Context)
Citation Context ...typical implementation of a simple simulation model show encouraging results in terms of simulation speedup and introduced error. 1 Introduction In traditional parallel discrete-event simulation (cf. =-=[2]-=-) the set of state variables of a simulation model is decomposed into subsets. Each of these is assigned to a logical process that manages the corresponding substate. The drawbacks of this approach ar... |

294 |
Parallel and Distributed Simulation Systems
- Fujimoto
- 2000
(Show Context)
Citation Context ...sses of models which exhibit complex states. Therefore, the concept of approximate state matching is proposed here as an efficient solution to the problem. 2 2.1 Time-Parallel Simulation According to =-=[4]-=-, a discrete event simulation essentially consists of the state variables, which establish the state space, the event list, which contains the events to happen in the future of the simulation, and the... |

39 |
Space-time and simulation
- Chandy, Sherman
- 1989
(Show Context)
Citation Context ...tates in the model. Time-parallel simulation takes a different approach by partitioning the time axis of an intended simulation execution and performing the simulation of these partitions in parallel =-=[1]-=-. Afterwards, the results of all partitions are combined to create the over1 all simulation result. This has the potential for massive parallelism, as the number of processes is determined by the numb... |

32 |
Algorithms for unboundedly parallel simulations
- Greenberg, Lubachevsky, et al.
- 1991
(Show Context)
Citation Context ...ect states. For the specific application of simulating cache accesses presented in [6], it is possible to do these computations without completely rerunning the simulation of the incorrect partition. =-=[5]-=- reduces the problem of simulation of certain kinds of queueing networks to recurrence relations that can be solved on massively parallel computers.sThe approaches mentioned above guarantee correct re... |

31 | Exploiting temporal uncertainty in parallel and distributed simulations
- Fujimoto
- 1999
(Show Context)
Citation Context ...pproaches in traditional space-parallel simulation which accept an error introduced by the simulation algorithm in order to reduce the overhead for synchronization between processes ([8], [10], [11], =-=[3]-=-). The challenge of the approach introduced here is to determine the rules for the allowed deviation of states for a given simulation model, such that the efficiency of the parallel simulation is maxi... |

29 |
Parallel trace-driven cache simulation by time partitioning
- Heidelberger, Stone
- 1990
(Show Context)
Citation Context ...t and continuing until the regeneration point is reached again. Afterwards the traces of the parallel simulations are concatenated to get a correct trace of the simulation over the whole time period. =-=[6]-=- uses fix-up computations to correct the simulation runs of those partitions of the time axis that are known to have started from incorrect states. For the specific application of simulating cache acc... |

28 |
A time-division algorithm for parallel simualtion. ACM Trans. on Modeling and Computer Simulation
- Lin, Lazowska
- 1991
(Show Context)
Citation Context ...ion quickly decreases with every additional simulation run of the same partition. Several different methods for reducing the cost of achieving state consistency between partitions have been proposed. =-=[7]-=- introduces the notion of regeneration points which are states that keep reoccuring throughout a simulation execution. If these points are known a priori, a number of simulations can be executed in pa... |

21 |
A Tutorial Introduction to Occam Programming
- Pountain, May
- 1987
(Show Context)
Citation Context ...a time parallel simulation with approximative state matching is given in Algorithm 1. As the algorithm describes a parallel simulation approach, a notation inspired by the Occam programming language (=-=[9]-=-) is used to denote parallel execution of statements. The designator 3 Figure 2 Approximative State Matching State Pk−1 ˆsk tk ˇsk−1 Pk AD(ˇsk−1) Time for all <list> do in parallel <statement> end for... |

17 | Unsynchronized parallel discrete event simulation
- Rao, Thondugulam, et al.
- 1998
(Show Context)
Citation Context ...e are also approaches in traditional space-parallel simulation which accept an error introduced by the simulation algorithm in order to reduce the overhead for synchronization between processes ([8], =-=[10]-=-, [11], [3]). The challenge of the approach introduced here is to determine the rules for the allowed deviation of states for a given simulation model, such that the efficiency of the parallel simulat... |

15 |
Tolerant synchronization for distributed simulations of interconnected computer networks
- Martini, Rümekasten, et al.
- 1997
(Show Context)
Citation Context ... There are also approaches in traditional space-parallel simulation which accept an error introduced by the simulation algorithm in order to reduce the overhead for synchronization between processes (=-=[8]-=-, [10], [11], [3]). The challenge of the approach introduced here is to determine the rules for the allowed deviation of states for a given simulation model, such that the efficiency of the parallel s... |

13 |
Approximate time-parallel simulation of queueing systems with losses
- Wang, Abrams
- 1992
(Show Context)
Citation Context ...ng to predefined rules. An obvious drawback of this approach is the error introduced by the approximate state matching. It must be considered that this error might even invalidate simulation results. =-=[12]-=- utilizes the concept of unmatched sets to weaken the state-match problem. In this paper, two system states partially match, if a subset of the state variables match. The unmatched set comprises the r... |

6 |
Time-parallel trace-driven simulation of CSMA/CD
- Wu, Fujimoto, et al.
- 2003
(Show Context)
Citation Context ...dle time can be minimized as long as there are partitions to simulate. Therefore, it is important to be able to provide unsimulated partitions to idle processes most of the time. As also discussed in =-=[13]-=-, the usage of a small number of partitions (e. g. equal to the number of processors) leads to an unbalanced processor load and consequently to a reduced efficiency of the overall simulation. Figure 4... |

3 | Relaxing causal constraints in pdes
- Thondugulam, Rao, et al.
- 1999
(Show Context)
Citation Context ...also approaches in traditional space-parallel simulation which accept an error introduced by the simulation algorithm in order to reduce the overhead for synchronization between processes ([8], [10], =-=[11]-=-, [3]). The challenge of the approach introduced here is to determine the rules for the allowed deviation of states for a given simulation model, such that the efficiency of the parallel simulation is... |