Speculative parallel simulation with an adaptive throttle scheme

Tay, Seng Chuan; Teo, Yong Meng; Kong, Siew Theng

doi:10.1145/268823.268909

Cited by 24 publications

(4 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In addition, we selected a GVT computation interval of 128 batches on both platforms, with a batch size of 24 events. Although the results are somewhat sensitive to the GVT interval (as a small GVT interval acts as a throttle to the simulation [Tay et al 1997]), these values are in the range where ROSS-MT is most efficient across a majority of the models.…”

Section: Pdes Slowdown Under Interferencementioning

confidence: 99%

Air

Wang

Abu-Ghazaleh

Ponomarev

2015

ACM Trans. Model. Comput. Simul.

View full text Add to dashboard Cite

Parallel discrete event simulation (PDES) harnesses parallel processing to improve the performance and capacity of simulation, supporting bigger and more detailed models simulated for more scenarios. The presence of interference from other users can lead to dramatic slowdown in the performance of the simulation. Interference is typically managed using operating system scheduling support (e.g., gang scheduling), a heavyweight approach with some drawbacks. We propose an application-level approach to interference resilience through alternative simulation scheduling and mapping algorithms. More precisely, the most resilient simulators allow dynamic mapping of simulation event execution to processing resources (a work pool model). However, this model has significant scheduling overhead and poor cache locality. Thus, we investigate using application-level interference mitigation where the application detects the presence of interference and reacts by changing the thread task allocation. Specifically, we propose a locality-aware adaptive dynamic mapping (LADM) algorithm that adjusts the number of active threads on the fly by detecting the presence of interference. LADM avoids having the application stall when threads are inactive due to context switching. We investigate different mechanisms for monitoring the level of interference and different approaches for remapping tasks. We show that LADM can substantially reduce the impact of interference while maintaining memory locality. ACM Reference Format:Jingjing Wang, Nael Abu-Ghazaleh, and Dmitry Ponomarev. 2015. AIR: Application-level interference resilience for PDES on multicore Systems. ACM Trans. Model.

show abstract

Section: Pdes Slowdown Under Interferencementioning

confidence: 99%

Air

Wang

Abu-Ghazaleh

Ponomarev

2015

ACM Trans. Model. Comput. Simul.

View full text Add to dashboard Cite

show abstract

“…In addition, we selected a GVT interval of 128 on both platforms, with a batch size of 24 events. Although the results are somewhat sensitive to the GVT interval (as a small GVT interval acts as a throttle to the simulation [26]), these values are in the range where ROSS-MT is most efficient across a range of models.…”

Section: Pdes Slowdown Under Interferencementioning

confidence: 99%

Interference resilient PDES on multi-core systems

Wang

Abu-Ghazaleh

Ponomarev

2013

Proceedings of the 1st ACM SIGSIM Conference on Principles of Advanced Discrete Simulation

View full text Add to dashboard Cite

Parallel Discrete Event Simulation (PDES) harnesses the power of parallel processing to improve the performance and capacity of simulation, supporting bigger models, in more details and for more scenarios. PDES engines are typically designed and evaluated assuming a homogeneous parallel computing system that is dedicated to the simulation application. In this paper, we first show that the presence of interference from other users, even a single process in an arbitrarily large parallel environment, can lead to dramatic slowdown in the performance of the simulation. We define a new metric, which we call proportional slowdown, that represents the idealized target for graceful slowdown in the presence of interference. We identify some of the reasons why simulators fall far short of proportional slowdown. Based on these observations, we design alternative simulation scheduling and mapping algorithms that are better able to tolerate interference. More precisely, the most resilient simulators will allow dynamic mapping of simulation event execution to processing resources (a work pool model). However, this model has significant overhead and can substantially impact locality. Thus, we propose a locality-aware adaptive dynamic-mapping (LADM) algorithm for PDES on multicore systems. LADM reduces the number of active threads in the presence of interference, avoiding having threads disabled due to context switching. We show that LADM can substantially reduce the impact of interference while maintaining memory locality reducing the gap with proportional slowdown. LADM and similar techniques can also help in situations where there is load imbalance or processor heterogeneity.

show abstract

“…The degree of optimism is then calculated to either block or process an event in a LP. Another approach proposed in [139] distinguishes between slower and faster LPs (in terms of progress in LVT) and determine the events to be processed on each LP. It restricts the progress of faster LPs and allows slower LPs to catch them.…”

Section: Adaptive Synchronisation In Pdesmentioning

confidence: 99%

Synchronised Range Queries

Suryanarayanan

Minson

Theodoropoulus

2009

2009 13th IEEE/ACM International Symposium on Distributed Simulation and Real Time Applications

View full text Add to dashboard Cite

Computer simulations have been used more than ever before to embark on developing and understanding complex systems such as Multi-Agent Systems (MAS). MAS are complex, non-deterministic, data-centric behaviour and nature. Simulations play a key role for the designer of an agent based system to experiment and study the impact of different architectures, environment and agent behaviour. As MAS are increasingly used to solve larger and more complex problems, scalability becomes an important issue for the successful deployment. An emerging viable solution is adopting distributed simulation techniques in executing MAS models in parallel. One approach is to distribute the shared state (or environment) of a simulation model across available computing resources. Based on this approach, PDES-MAS (Parallel and Discrete Event Simulations for Multi-Agent System) framework is designed to execute large scale models such as MAS. It adopts PDES techniques to distribute and run parallel simulation of Multi-Agent Systems (MAS). Several challenges arise on executing MAS models on a distributed environment, of which one issue that requires focus is data access. Accessing data efficiently in a latency-sensitive and large scale network overlay is a vital requirement for the scalability of the system. Following PDES paradigm, it is also very important that events (accessing shared state) in a parallel and discrete event simulation system are processed in a non decreasing (logical) time stamp order. So, this thesis presents a notion of logical time synchronised range queries to access data and in particular within the PDES-MAS framework. To localise data access, this thesis also provides mechanisms to distribute shared state in an adaptive manner such that the distribution reflects access patterns of simulating nodes. The algorithms are evaluated within the implementation of PDES-MAS framework using various agent based simulation traces. ACKNOWLEDGEMENTS

show abstract

Speculative parallel simulation with an adaptive throttle scheme

Cited by 24 publications

References 12 publications

Air

Air

Interference resilient PDES on multi-core systems

Synchronised Range Queries

Contact Info

Product

Resources

About