Language based state saving extensions for optimistic parallel simulation

Gomes, Fabian; Unger, Brian; Cleary, John G.

doi:10.1145/256562.256812

Cited by 4 publications

(3 citation statements)

References 16 publications

(2 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Time Warp needs to overcome several problems in order to maintain good efficiency. These problems have prompted a flood of research in areas of state saving (e.g., Gomes et al, 1996;Lin and Lazowska, 1990;Lin et al, 1993;Ronngren et al, 1996), rollback (e.g., Gafni, 1988;Reiher et al, 1990;West, 1988), GVT computation (e.g., Fujimoto and Hybinette, 1997;Mattern, 1993;Samadi, 1985), memory management (e.g., Jefferson, 1990;Lin and Preiss, 1991;Preiss and Loucks, 1995), and alternative optimistic execution (e.g., Dickens and Reynolds, 1990;Sokol et al, 1988;Steinman, 1991Steinman, , 1993. The jury is out on which of the two approaches is a better choice.…”

Section: Parallel and Distributed Simulationmentioning

confidence: 99%

Parallel and Distributed Immersive Real-Time Simulation of Large-Scale Networks

Liu¹

2010

Parallel and Distributed Computing

View full text Add to dashboard Cite

Network researchers need to embrace the challenge of designing the next-generation highperformance networking and software infrastructures that address the growing demand of distributed applications. These applications, particularly those potential "game changers" or "killer apps", such as voice-over-IP (VoIP) and peer-to-peer (P2P) applications surfaced in recent years, will significantly influence the way people conduct business and go about their daily lives. These distributed applications also include platforms that facilitate large-scale scientific experimentation through remote control and visualization. Many large-scale science applications-such as those in the field of astronomy, astrophysics, climate and environmental science, material science, particle physics, and social science-depend on the availability of high-performance facilities and advanced experimental instruments. Extreme networking capabilities together with effective high-end middleware infrastructures are of great importance to interconnecting these applications, computing resources and experimental facilities. "When all you have is a hammer, everything looks like a nail." The success of advancing critical technologies, to a large extent, depends on the available tools that can help effectively prototype, test, and analyze new designs and new ideas. Traditionally, network research has relied on a variety of tools. Physical network testbeds, such as WAIL (Barford and Landweber, 2003) and PlanetLab (Peterson et al., 2002), provide physical network connectivity; these testbeds are designed specifically for studying network protocols and services under real network conditions. However, the network condition of these testbeds is by and large constrained by the physical setup of the system and therefore inflexible for network researchers to explore a wide spectrum of the design space. To allow more flexibility, some of these testbeds, such as EmuLab (White et al., 2002) and VINI (Bavier et al., 2006), also offer emulation capabilities by modulating network traffic according to configuration and traffic condition of the target network. Physical and emulation testbeds currently are the mainstream for experimental networking research, primarily due to their capability of achieving desirable realism and accuracy. These testbeds, however, are costly to build. Due to limited resources available, conducting prolonged large-scale experiments on these platforms is difficult. Another solution is to use analytical models. Although analytical models are capable of bringing us important insight to the system design, dealing with a system as complex as the global network requires significantly simplified assumptions to be made to keep the models tractable. These simplified assumptions often 12 www.intechopen.com

show abstract

Section: Parallel and Distributed Simulationmentioning

confidence: 99%

Parallel and Distributed Immersive Real-Time Simulation of Large-Scale Networks

Liu¹

2010

Parallel and Distributed Computing

View full text Add to dashboard Cite

show abstract

“…Various methods for efficient checkpointing of objects have been proposed. These include copy state-saving -complete state is saved, incremental state-saving -only modified state is saved (Steinman 1993), hardware-based state-saving (Fujimoto, Tsai, and Gop-alakrishnan 1988), and language extensions for state-saving (Gomes, Unger, and Cleary 1996). Efficient schemes for incremental state-saving, such as those employed by Steinmain (1993)) require user-intervention for checkpointing and state-saving.…”

Section: Related Workmentioning

confidence: 99%

Checkpoint and recovery methods in the ParaSol simulation system

Mascarenhas¹,

Knop²,

Pasquini³

et al. 1997

Proceedings of the 29th Conference on Winter Simulation - WSC '97

View full text Add to dashboard Cite

State-saving operations are a major source of overheads in optimistic and adaptive parallel discreteevent simulations.We present some techniques for saving state in the the context of the PARASOL multithreaded parallel simulation system. In this system, threads are used to implement both logical processes and active transactions which access passive simulation objects. Hence, system state is a combination of thread-state and object-state. We introduce a new save-if-modified method for incremental checkpointing of threads and objects. Because of the PARA-SOL system's domain-oriented support, checkpointing is transparent to the user. Application-level objects that are foreign to a domain may be saved via invocations to primitives in the system's ParaState module.

show abstract

“…We consider that optimistic synchronization methods may not be adapted to this kind of simulation, for two main reasons. First, because of the complexity of the components present in network models, the necessary state-saving operations would lead to a significant overhead, even if some recent techniques do optimize these operations [7]. The second reason is related to stability and implementation costs: the additional code achieving state-saving may be complex to write, and implementation errors would lead to potentially unnoticed inconsistencies in simulation results.…”

Section: Introductionmentioning

confidence: 99%

An empirical study of different strategies for the parallel simulation of large-scale communication networks

Forgeau

Killat

Proceedings International Parallel and Distributed Processing Symposium

View full text Add to dashboard Cite

An MPI-based parallel simulation system for Wide Area Networks is described in this article, as well as some characteristics of the models it handles. A simple synchronization algorithm is then presented along with some additional synchronization strategies that could improve the performance. These strategies are tested on a reference model, what shows how the simulator scales and how it is sensitive to parameters such as connectivity and send granularity.

show abstract

Language based state saving extensions for optimistic parallel simulation

Abstract: One of the greatest challenges in making optimistic synchronization techniques such as Time Warp practical tools is making state saving efficient and easy to use, State saving is necessary so that when optimistic execution is found to be out of order, rollback can be

Cited by 4 publications

References 16 publications

Parallel and Distributed Immersive Real-Time Simulation of Large-Scale Networks

Parallel and Distributed Immersive Real-Time Simulation of Large-Scale Networks

Checkpoint and recovery methods in the ParaSol simulation system

An empirical study of different strategies for the parallel simulation of large-scale communication networks

Contact Info

Product

Resources

About