ACM Transactions on Programming Languages and Systems (TOPLAS)
Limitation of optimism in the time warp operating system
WSC '89 Proceedings of the 21st conference on Winter simulation
Parallel discrete event simulation
Communications of the ACM - Special issue on simulation
An analytical comparison of periodic checkpointing and incremental state saving
PADS '93 Proceedings of the seventh workshop on Parallel and distributed simulation
Adaptive checkpointing in Time Warp
PADS '94 Proceedings of the eighth workshop on Parallel and distributed simulation
Comparative analysis of periodic state saving techniques in time warp simulators
PADS '95 Proceedings of the ninth workshop on Parallel and distributed simulation
Software architecture: perspectives on an emerging discipline
Software architecture: perspectives on an emerging discipline
A framework for performance analysis of parallel discrete event simulators
Proceedings of the 29th conference on Winter simulation
Optimizing communication in time-warp simulators
PADS '98 Proceedings of the twelfth workshop on Parallel and distributed simulation
Adaptive Control
External Adjustment of Runtime Parameters in Time Warp Synchronized Parallel Simulators
IPPS '97 Proceedings of the 11th International Symposium on Parallel Processing
| Hi-index | 0.00 |
Parallel simulations using optimistic synchronization strategies such as Time Warp, operate with no regard to global synchronization since this results in greater parallelism and lower synchronization cost. However, like virtual memory, the parallel simulators may end up thrashing instead of performing useful work. The complication in using a Time Warp simulator is then to configure it suitably for good performance and avoid thrashing. Unfortunately, the optimal configuration is not generally static among different applications or even throughout an entire run of a single application. Thus, online feedback control systems are deployed to govern the adjustment of input parameters in our Time Warp simulation kernel. The design and implementation of effective feedback control systems can be difficult; the extra processing is pure overhead that must be absorbed by any performance gains delivered. The problem is further complicated when attempting to build a simulation kernel that is designed efficiently to operate with many different applications. In this paper, we introduce a control-centric architecture that is used to monitor and manage different parts of a Time Warp simulator. Specifically, we extend concepts from control theory such as adaptive control and stability, to better understand and design hierarchically-distributed run-time control systems for Time Warp based parallel simulation.